text-classification bool 2 classes | text stringlengths 0 664k |
|---|---|
false |
# Dataset Card for BrWaC
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Homepage:** [BrWaC homepage](https://www.inf.ufrgs.br/pln/wiki/index.php?title=BrWaC)
- **Repository:** [BrWaC repository](https://www.inf.ufrgs.br/pln/wiki/index.php?title=BrWaC)
- **Paper:** [The brWaC Corpus: A New Open Resource for Brazilian Portuguese](https://www.aclweb.org/anthology/L18-1686/)
- **Point of Contact:** [Jorge A. Wagner Filho](mailto:jawfilho@inf.ufrgs.br)
### Dataset Summary
The BrWaC (Brazilian Portuguese Web as Corpus) is a large corpus constructed following the Wacky framework,
which was made public for research purposes. The current corpus version, released in January 2017, is composed by
3.53 million documents, 2.68 billion tokens and 5.79 million types. Please note that this resource is available
solely for academic research purposes, and you agreed not to use it for any commercial applications. No need to manually download external sources.
### Supported Tasks and Leaderboards
[More Information Needed]
### Languages
Portuguese
## Dataset Structure
### Data Instances
An example from the BrWaC dataset looks as follows:
```
{
"doc_id": "netg-1afc73",
"text": {
"paragraphs": [
[
"Conteúdo recente"
],
[
"ESPUMA MARROM CHAMADA \"NINGUÉM MERECE\""
],
[
"31 de Agosto de 2015, 7:07 , por paulo soavinski - | No one following this article yet."
],
[
"Visualizado 202 vezes"
],
[
"JORNAL ELETRÔNICO DA ILHA DO MEL"
],
[
"Uma espuma marrom escuro tem aparecido com frequência na Praia de Fora.",
"Na faixa de areia ela aparece disseminada e não chama muito a atenção.",
"No Buraco do Aipo, com muitas pedras, ela aparece concentrada.",
"É fácil saber que esta espuma estranha está lá, quando venta.",
"Pequenos algodões de espuma começam a flutuar no espaço, pertinho da Praia do Saquinho.",
"Quem pode ajudar na coleta deste material, envio a laboratório renomado e pagamento de análises, favor entrar em contato com o site."
]
]
},
"title": "ESPUMA MARROM CHAMADA ‟NINGUÃÂM MERECE‟ - paulo soavinski",
"uri": "http://blogoosfero.cc/ilhadomel/pousadasilhadomel.com.br/espuma-marrom-chamada-ninguem-merece"
}
```
### Data Fields
- `doc_id`: The document ID
- `title`: The document title
- `uri`: URI where the document was extracted from
- `text`: A list of document paragraphs (with a list of sentences in it as a list of strings)
### Data Splits
The data is only split into train set with size of 3530796 samples.
## Dataset Creation
### Curation Rationale
[More Information Needed]
### Source Data
#### Initial Data Collection and Normalization
[More Information Needed]
#### Who are the source language producers?
[More Information Needed]
### Annotations
#### Annotation process
[More Information Needed]
#### Who are the annotators?
[More Information Needed]
### Personal and Sensitive Information
[More Information Needed]
## Considerations for Using the Data
### Social Impact of Dataset
[More Information Needed]
### Discussion of Biases
[More Information Needed]
### Other Known Limitations
[More Information Needed]
## Additional Information
### Dataset Curators
[More Information Needed]
### Licensing Information
[More Information Needed]
### Citation Information
```
@inproceedings{wagner2018brwac,
title={The brwac corpus: A new open resource for brazilian portuguese},
author={Wagner Filho, Jorge A and Wilkens, Rodrigo and Idiart, Marco and Villavicencio, Aline},
booktitle={Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)},
year={2018}
}
``` |
false |
# Wikipedia (ar) embedded with cohere.ai `multilingual-22-12` encoder
We encoded [Wikipedia (ar)](https://ar.wikipedia.org) using the [cohere.ai](https://txt.cohere.ai/multilingual/) `multilingual-22-12` embedding model.
To get an overview how this dataset was created and pre-processed, have a look at [Cohere/wikipedia-22-12](https://huggingface.co/datasets/Cohere/wikipedia-22-12).
## Embeddings
We compute for `title+" "+text` the embeddings using our `multilingual-22-12` embedding model, a state-of-the-art model that works for semantic search in 100 languages. If you want to learn more about this model, have a look at [cohere.ai multilingual embedding model](https://txt.cohere.ai/multilingual/).
## Further languages
We provide embeddings of Wikipedia in many different languages:
[ar](https://huggingface.co/datasets/Cohere/wikipedia-22-12-ar-embeddings), [de](https://huggingface.co/datasets/Cohere/wikipedia-22-12-de-embeddings), [en](https://huggingface.co/datasets/Cohere/wikipedia-22-12-en-embeddings), [es](https://huggingface.co/datasets/Cohere/wikipedia-22-12-es-embeddings), [fr](https://huggingface.co/datasets/Cohere/wikipedia-22-12-fr-embeddings), [hi](https://huggingface.co/datasets/Cohere/wikipedia-22-12-hi-embeddings), [it](https://huggingface.co/datasets/Cohere/wikipedia-22-12-it-embeddings), [ja](https://huggingface.co/datasets/Cohere/wikipedia-22-12-ja-embeddings), [ko](https://huggingface.co/datasets/Cohere/wikipedia-22-12-ko-embeddings), [simple english](https://huggingface.co/datasets/Cohere/wikipedia-22-12-simple-embeddings), [zh](https://huggingface.co/datasets/Cohere/wikipedia-22-12-zh-embeddings),
You can find the Wikipedia datasets without embeddings at [Cohere/wikipedia-22-12](https://huggingface.co/datasets/Cohere/wikipedia-22-12).
## Loading the dataset
You can either load the dataset like this:
```python
from datasets import load_dataset
docs = load_dataset(f"Cohere/wikipedia-22-12-ar-embeddings", split="train")
```
Or you can also stream it without downloading it before:
```python
from datasets import load_dataset
docs = load_dataset(f"Cohere/wikipedia-22-12-ar-embeddings", split="train", streaming=True)
for doc in docs:
docid = doc['id']
title = doc['title']
text = doc['text']
emb = doc['emb']
```
## Search
A full search example:
```python
#Run: pip install cohere datasets
from datasets import load_dataset
import torch
import cohere
co = cohere.Client(f"<<COHERE_API_KEY>>") # Add your cohere API key from www.cohere.com
#Load at max 1000 documents + embeddings
max_docs = 1000
docs_stream = load_dataset(f"Cohere/wikipedia-22-12-ar-embeddings", split="train", streaming=True)
docs = []
doc_embeddings = []
for doc in docs_stream:
docs.append(doc)
doc_embeddings.append(doc['emb'])
if len(docs) >= max_docs:
break
doc_embeddings = torch.tensor(doc_embeddings)
query = 'Who founded Youtube'
response = co.embed(texts=[query], model='multilingual-22-12')
query_embedding = response.embeddings
query_embedding = torch.tensor(query_embedding)
# Compute dot score between query embedding and document embeddings
dot_scores = torch.mm(query_embedding, doc_embeddings.transpose(0, 1))
top_k = torch.topk(dot_scores, k=3)
# Print results
print("Query:", query)
for doc_id in top_k.indices[0].tolist():
print(docs[doc_id]['title'])
print(docs[doc_id]['text'], "\n")
```
## Performance
You can find performance on the MIRACL dataset (a semantic search evaluation dataset) here: [miracl-en-queries-22-12#performance](https://huggingface.co/datasets/Cohere/miracl-en-queries-22-12#performance) |
false | This is the reporsitory of Turkish fake news dataset which consists of Zaytung posts and Hurriyet news articles.
Code folder contains the web scrapper python files.
Raw folder contains txt files downloaded from sources.
Clean folder contains txt files in lowercase, punctuation and numbers removed. |
false |
# Dataset Card for [Dataset Name]
## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Point of Contact:** [Vishal Burman](mailto:vishal.a.burman23@gmail.com)
### Dataset Summary
This dataset comprises of open-domain question-answer pairs obtained from extracting 150K FAQ URLs from C4 dataset. Please refer to the original [`paper`](https://arxiv.org/abs/1910.10683) and [`dataset card`](https://huggingface.co/datasets/c4) for more details.
You can load C4-FAQs as follows:
```python
from datasets import load_dataset
c4_faqs_dataset = load_dataset("vishal-burman/c4-faqs")
```
### Supported Tasks and Leaderboards
C4-FAQs is mainly intended for open-domain end-to-end question generation. It can also be used for open-domain question answering.
### Languages
C4-FAQs only supports English language.
## Dataset Structure
### Data Instances
An example of a single dataset point:
```python
{'url': 'https://www.brusselsghosts.com/things-to-do-brussels/faq.html', 'faq_pairs': [{'question': 'What should I bring for the tour?', 'answer': 'Nothing special, just be ready to walk for bit and potentially something to protect you from poltergeists and rain. Any kind of amulet or protection stone is also welcome.'}, {'question': 'Can kids join too ?', 'answer': 'Yes, we accept kids from 6 years old and on! We also have a family discount, if you book for 2 adults and 2 kids!'}, {'question': 'Where is the meeting point ?', 'answer': 'Brussels has many paved roads and those are hardly accessible with a wheelchair, for that reason we have to unfortunately label our tour as not wheelchair accessible.'}]}
```
### Data Fields
The data have several fields:
- `url`: URL of the webpage containing the FAQs
- `faq_pairs`: A list of question-answer pairs extracted from the webpage
- `question`: A single question as a string
- `answer`: A single answer to the above question as a string
### Data Splits
| subset | total |
|:-------|:------|
| train | 150K |
## Dataset Creation
### Curation Rationale
The dataset was curated to create end-to-end Question Generation pipelines. A large amount of open-source models use [`SQuAD`](https://huggingface.co/datasets/squad) dataset to create answer-agnostic question generation models. While the questions are valid, they often are short factoid in nature. This dataset is curated from FAQs of websites, which are generally hand-crafted and can be used to further improve generated question quality.
## Additional Information
### Dataset Curators
Original data by [Common Crawl](https://commoncrawl.org/).
### Licensing Information
The original dataset was released under the terms of ODC-BY. By using this, you are also bound by the Common Crawl terms of use in respect of the content contained in the dataset.
### Citation Information
If you use this dataset, I would love to hear about it! Reach out on GitHub, twitter or shoot me an email.
To cite the original `c4` dataset:
```bibtex
@article{2019t5,
author = {Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu},
title = {Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer},
journal = {arXiv e-prints},
year = {2019},
archivePrefix = {arXiv},
eprint = {1910.10683},
}
```
|
false |
# Датасет для задачи раскрытия неполных реплик в контексте диалога
Подробное описание задачи "Incomplete Utterance Restoration" можно найти в карточке генеративной модели [inkoziev/rugpt_interpreter](https://huggingface.co/inkoziev/rugpt_interpreter), которая обучена на аугментированном варианте этого датасета.
В датасете содержатся фрагменты диалогов длиной от 1 до 3 последовательных реплик. Для последней реплики дается ее полный вариант с раскрытыми анафорами, эллипсисами и т.д.
Например, следующий сэмпл:
```
{
"context": ["Добрый день, давай поговорим?"],
"short_phrase": "добрый день, давай",
"expanded_phrase": "Давай поговорим"
}
```
Соответствует фрагменту диалога:
```
- Добрый день, давай поговорим?
- добрый день, давай
```
Для последней реплики "_добрый день, давай_" приводится ее полный вариант "_Давай поговорим_".
Для фрагментов длиной 3 реплики в поле ```context``` будет 2 элемента:
```
{
"context": [
"Что-то зрителей полный инет, а изучителей не видать.",
"Ты кем себя считаешь: зрителем или изучателем?"
],
"short_phrase": "изучателем",
"expanded_phrase": "я считаю себя изучателем"
},
```
Этот сэмпл соответствует фрагменту диалога:
```
- Что-то зрителей полный инет, а изучителей не видать.
- Ты кем себя считаешь: зрителем или изучателем?
- изучателем
```
Для последней реплики дана полная форма "_я считаю себя изучателем_".
## Особенности датасета
1) Для многих сэмплов вежливая форма 2л. мн.ч. "Вы" в раскрытом варианте нормализована до "ты".
2) Есть нецензурная лексика, оскорбления и т.п.
## Ссылки, цитирование
Этот датасет я собирал сам для экспериментов с [диалоговыми системами](https://github.com/Koziev/chatbot). Если вы используете его в каком-то виде, буду благодарен за ссылки на эту страницу.
|
false |
# MIRACL (ar) embedded with cohere.ai `multilingual-22-12` encoder
We encoded the [MIRACL dataset](https://huggingface.co/miracl) using the [cohere.ai](https://txt.cohere.ai/multilingual/) `multilingual-22-12` embedding model.
The query embeddings can be found in [Cohere/miracl-ar-queries-22-12](https://huggingface.co/datasets/Cohere/miracl-ar-queries-22-12) and the corpus embeddings can be found in [Cohere/miracl-ar-corpus-22-12](https://huggingface.co/datasets/Cohere/miracl-ar-corpus-22-12).
For the orginal datasets, see [miracl/miracl](https://huggingface.co/datasets/miracl/miracl) and [miracl/miracl-corpus](https://huggingface.co/datasets/miracl/miracl-corpus).
Dataset info:
> MIRACL 🌍🙌🌏 (Multilingual Information Retrieval Across a Continuum of Languages) is a multilingual retrieval dataset that focuses on search across 18 different languages, which collectively encompass over three billion native speakers around the world.
>
> The corpus for each language is prepared from a Wikipedia dump, where we keep only the plain text and discard images, tables, etc. Each article is segmented into multiple passages using WikiExtractor based on natural discourse units (e.g., `\n\n` in the wiki markup). Each of these passages comprises a "document" or unit of retrieval. We preserve the Wikipedia article title of each passage.
## Embeddings
We compute for `title+" "+text` the embeddings using our `multilingual-22-12` embedding model, a state-of-the-art model that works for semantic search in 100 languages. If you want to learn more about this model, have a look at [cohere.ai multilingual embedding model](https://txt.cohere.ai/multilingual/).
## Loading the dataset
In [miracl-ar-corpus-22-12](https://huggingface.co/datasets/Cohere/miracl-ar-corpus-22-12) we provide the corpus embeddings. Note, depending on the selected split, the respective files can be quite large.
You can either load the dataset like this:
```python
from datasets import load_dataset
docs = load_dataset(f"Cohere/miracl-ar-corpus-22-12", split="train")
```
Or you can also stream it without downloading it before:
```python
from datasets import load_dataset
docs = load_dataset(f"Cohere/miracl-ar-corpus-22-12", split="train", streaming=True)
for doc in docs:
docid = doc['docid']
title = doc['title']
text = doc['text']
emb = doc['emb']
```
## Search
Have a look at [miracl-ar-queries-22-12](https://huggingface.co/datasets/Cohere/miracl-ar-queries-22-12) where we provide the query embeddings for the MIRACL dataset.
To search in the documents, you must use **dot-product**.
And then compare this query embeddings either with a vector database (recommended) or directly computing the dot product.
A full search example:
```python
# Attention! For large datasets, this requires a lot of memory to store
# all document embeddings and to compute the dot product scores.
# Only use this for smaller datasets. For large datasets, use a vector DB
from datasets import load_dataset
import torch
#Load documents + embeddings
docs = load_dataset(f"Cohere/miracl-ar-corpus-22-12", split="train")
doc_embeddings = torch.tensor(docs['emb'])
# Load queries
queries = load_dataset(f"Cohere/miracl-ar-queries-22-12", split="dev")
# Select the first query as example
qid = 0
query = queries[qid]
query_embedding = torch.tensor(queries['emb'])
# Compute dot score between query embedding and document embeddings
dot_scores = torch.mm(query_embedding, doc_embeddings.transpose(0, 1))
top_k = torch.topk(dot_scores, k=3)
# Print results
print("Query:", query['query'])
for doc_id in top_k.indices[0].tolist():
print(docs[doc_id]['title'])
print(docs[doc_id]['text'])
```
You can get embeddings for new queries using our API:
```python
#Run: pip install cohere
import cohere
co = cohere.Client(f"{api_key}") # You should add your cohere API Key here :))
texts = ['my search query']
response = co.embed(texts=texts, model='multilingual-22-12')
query_embedding = response.embeddings[0] # Get the embedding for the first text
```
## Performance
In the following table we compare the cohere multilingual-22-12 model with Elasticsearch version 8.6.0 lexical search (title and passage indexed as independent fields). Note that Elasticsearch doesn't support all languages that are part of the MIRACL dataset.
We compute nDCG@10 (a ranking based loss), as well as hit@3: Is at least one relevant document in the top-3 results. We find that hit@3 is easier to interpret, as it presents the number of queries for which a relevant document is found among the top-3 results.
Note: MIRACL only annotated a small fraction of passages (10 per query) for relevancy. Especially for larger Wikipedias (like English), we often found many more relevant passages. This is know as annotation holes. Real nDCG@10 and hit@3 performance is likely higher than depicted.
| Model | cohere multilingual-22-12 nDCG@10 | cohere multilingual-22-12 hit@3 | ES 8.6.0 nDCG@10 | ES 8.6.0 acc@3 |
|---|---|---|---|---|
| miracl-ar | 64.2 | 75.2 | 46.8 | 56.2 |
| miracl-bn | 61.5 | 75.7 | 49.2 | 60.1 |
| miracl-de | 44.4 | 60.7 | 19.6 | 29.8 |
| miracl-en | 44.6 | 62.2 | 30.2 | 43.2 |
| miracl-es | 47.0 | 74.1 | 27.0 | 47.2 |
| miracl-fi | 63.7 | 76.2 | 51.4 | 61.6 |
| miracl-fr | 46.8 | 57.1 | 17.0 | 21.6 |
| miracl-hi | 50.7 | 62.9 | 41.0 | 48.9 |
| miracl-id | 44.8 | 63.8 | 39.2 | 54.7 |
| miracl-ru | 49.2 | 66.9 | 25.4 | 36.7 |
| **Avg** | 51.7 | 67.5 | 34.7 | 46.0 |
Further languages (not supported by Elasticsearch):
| Model | cohere multilingual-22-12 nDCG@10 | cohere multilingual-22-12 hit@3 |
|---|---|---|
| miracl-fa | 44.8 | 53.6 |
| miracl-ja | 49.0 | 61.0 |
| miracl-ko | 50.9 | 64.8 |
| miracl-sw | 61.4 | 74.5 |
| miracl-te | 67.8 | 72.3 |
| miracl-th | 60.2 | 71.9 |
| miracl-yo | 56.4 | 62.2 |
| miracl-zh | 43.8 | 56.5 |
| **Avg** | 54.3 | 64.6 |
|
false | # Dataset for project: food-category-classification-v2.0
## Dataset Description
This dataset for project food-category-classification-v2.0 was scraped with the help of a bulk google image downloader.
## Dataset Structure
### Dataset Fields
The dataset has the following fields (also called "features"):
```json
{
"image": "Image(decode=True, id=None)",
"target": "ClassLabel(names=['Bread', 'Dairy', 'Dessert', 'Egg', 'Fried Food', 'Fruit', 'Meat', 'Noodles', 'Rice', 'Seafood', 'Soup', 'Vegetable'], id=None)"
}
```
### Dataset Splits
This dataset is split into a train and validation split. The split sizes are as follows:
| Split name | Num samples |
| ------------ | ------------------- |
| train | 1200 |
| valid | 300 |
|
false | # Dataset Card for Cocktail Recipes
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
## Dataset Description
### Dataset Summary
Cocktail Recipes Dataset for Semi-Structured Text Generation.
### Supported Tasks and Leaderboards
[More Information Needed]
### Languages
The dataset is in English.
## Dataset Structure
### Data Instances
```json
{"title": "Final Ward",
"ingredients": ["0.75 oz. Rye Whiskey",
"0.75 oz. Lemon Juice",
"0.75 oz. Maraschino Liqueur",
"0.75 oz. Green Chartreuse"],
"directions": ["shake on ice and strain"],
"misc":[],
"source": "Death & Co.",
"ner":["whiskey",
"chartreuse",
"maraschino liqueur"]}
```
### Data Fields
- `title` (`str`): Title of the recipe.
- `ingredients` (`list` of `str`): Ingredients.
- `directions` (`list` of `str`): Instruction steps.
- `source` (`str`): Origin of each recipe
- `ner` (`list` of `str`): NER entities.
### Data Splits
The dataset contains a single `train` split.
## Dataset Creation
[More Information Needed]
### Curation Rationale
[More Information Needed]
### Source Data
[More Information Needed]
#### Initial Data Collection and Normalization
[More Information Needed]
#### Who are the source language producers?
[More Information Needed]
### Annotations
[More Information Needed]
#### Annotation process
[More Information Needed]
#### Who are the annotators?
[More Information Needed]
### Personal and Sensitive Information
[More Information Needed]
## Considerations for Using the Data
### Social Impact of Dataset
[More Information Needed]
### Discussion of Biases
[More Information Needed]
### Other Known Limitations
[More Information Needed]
## Additional Information
### Dataset Curators
[More Information Needed]
|
false |
# VGGSound
VGG-Sound is an audio-visual correspondent dataset consisting of short clips of audio sounds, extracted from videos uploaded to YouTube.
- **Homepage:** https://www.robots.ox.ac.uk/~vgg/data/vggsound/
- **Paper:** https://arxiv.org/abs/2004.14368
- **Github:** https://github.com/hche11/VGGSound
## Analysis
- **310+ classes:** VGG-Sound contains audios spanning a large number of challenging acoustic environments and noise characteristics of real applications.
- **200,000+ videos:** All videos are captured "in the wild" with audio-visual correspondence in the sense that the sound source is visually evident.
- **550+ hours:** VGG-Sound consists of both audio and video. Each segment is 10 seconds long.
## Download
We provide a csv file. For each YouTube video, we provide YouTube URLs, time stamps, audio labels and train/test split. Each line in the csv file has columns defined by here.
```
# YouTube ID, start seconds, label, train/test split.
```
And you can download VGGSound directly from this [repository](https://huggingface.co/datasets/Loie/VGGSound/tree/main).
## License
The VGG-Sound dataset is available to download for commercial/research purposes under a Creative Commons Attribution 4.0 International License. The copyright remains with the original owners of the video. A complete version of the license can be found [here](https://thor.robots.ox.ac.uk/datasets/vggsound/license_vggsound.txt).
## Citation
Please cite the following if you make use of the dataset.
```
@InProceedings{Chen20,
author = "Honglie Chen and Weidi Xie and Andrea Vedaldi and Andrew Zisserman",
title = "VGGSound: A Large-scale Audio-Visual Dataset",
booktitle = "International Conference on Acoustics, Speech, and Signal Processing (ICASSP)",
year = "2020",
}
``` |
false |
# Dataset Card for WikiTableQuestions-wtq
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-instances)
- [Data Splits](#data-instances)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
## Dataset Description
- **Homepage:** [WikiTableQuestions homepage](https://nlp.stanford.edu/software/sempre/wikitable)
- **Repository:** [WikiTableQuestions repository](https://github.com/ppasupat/WikiTableQuestions)
- **Paper:** [Compositional Semantic Parsing on Semi-Structured Tables](https://arxiv.org/abs/1508.00305)
- **Leaderboard:** [WikiTableQuestions leaderboard on PaperWithCode](https://paperswithcode.com/dataset/wikitablequestions)
- **Point of Contact:** [Needs More Information]
### Dataset Summary
The WikiTableQuestions-wtq dataset is a small-scale dataset for the task of question answering on semi-structured tables.
This data includes the `aggregation_label` and `answer_coordinates` to make it easy to train this model on any [TAPAS](https://huggingface.co/docs/transformers/model_doc/tapas#usage-finetuning) based modles.
### Supported Tasks and Leaderboards
question-answering, table-question-answering
### Languages
en
## Dataset Structure
### Data Instances
#### default
- **Size of downloaded dataset files:** 27.91 MB
- **Size of the generated dataset:** 45.68 MB
- **Total amount of disk used:** 73.60 MB
An example of 'validation' looks as follows:
```
{
"id": "nt-0",
"question": "What is the total average attendance at all USL First Division matches?",
"answers": [
"36755"
],
"table": {
"header": [
"Year",
"Division",
"League",
"Regular Season",
"Playoffs",
"Open Cup",
"Avg. Attendance"
],
"rows": [
[
"2001",
"2",
"USL A-League",
"4th, Western",
"Quarterfinals",
"Did not qualify",
"7,169"
],
...
],
"name": "csv/204-csv/590.tsv"
},
"aggregation_label": "SUM",
"answer_coordinates": [
[
4,
6
],
...
]
}
```
### Data Fields
The data fields are the same among all splits.
#### default
- `id`: a `string` feature.
- `question`: a `string` feature.
- `answers`: a `list` of `string` feature.
- `answers_coordinates`: a `list` of `int,int` tuples.
- `aggregation_label`: a `string` feature.
- `table`: a dictionary feature containing:
- `header`: a `list` of `string` features.
- `rows`: a `list` of `list` of `string` features:
- `name`: a `string` feature.
### Data Splits
TBA
## Dataset Creation
### Curation Rationale
[Needs More Information]
### Source Data
#### Initial Data Collection and Normalization
[Needs More Information]
#### Who are the source language producers?
[Needs More Information]
### Annotations
#### Annotation process
[Needs More Information]
#### Who are the annotators?
[Needs More Information]
### Personal and Sensitive Information
[Needs More Information]
## Considerations for Using the Data
### Social Impact of Dataset
[Needs More Information]
### Discussion of Biases
[Needs More Information]
### Other Known Limitations
[Needs More Information]
## Additional Information
### Dataset Curators
Panupong Pasupat and Percy Liang
### Licensing Information
Creative Commons Attribution Share Alike 4.0 International
### Citation Information
```
@inproceedings{pasupat-liang-2015-compositional,
title = "Compositional Semantic Parsing on Semi-Structured Tables",
author = "Pasupat, Panupong and Liang, Percy",
booktitle = "Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)",
month = jul,
year = "2015",
address = "Beijing, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/P15-1142",
doi = "10.3115/v1/P15-1142",
pages = "1470--1480",
}
```
### Contributions
Thanks to [@SivilTaram](https://github.com/SivilTaram) for adding this dataset. |
false | # Machine translation dataset for NLU (Virual Assistant) with slot transfer between languages
## Dataset Summary
Disclaimer: This is for research purposes only. Please have a look at the license section below. Some of the datasets used to construct IVA_MT have an unknown license.
IVA_MT is a machine translation dataset that can be used to train, adapt and evaluate MT models used in Virtual Assistant NLU context (e.g. to translate trainig corpus of NLU).
## Dataset Composition
### en-pl
| Corpus | Train | Dev | Test |
|----------------------------------------------------------------------|--------|-------|-------|
| [Massive 1.1](https://huggingface.co/datasets/AmazonScience/massive) | 11514 | 2033 | 2974 |
| [Leyzer 0.2.0](https://github.com/cartesinus/leyzer/tree/0.2.0) | 3974 | 701 | 1380 |
| [OpenSubtitles from OPUS](https://opus.nlpl.eu/OpenSubtitles-v1.php) | 2329 | 411 | 500 |
| [KDE from OPUS](https://opus.nlpl.eu/KDE4.php) | 1154 | 241 | 241 |
| [CCMatrix from Opus](https://opus.nlpl.eu/CCMatrix.php) | 1096 | 232 | 237 |
| [Ubuntu from OPUS](https://opus.nlpl.eu/Ubuntu.php) | 281 | 60 | 59 |
| [Gnome from OPUS](https://opus.nlpl.eu/GNOME.php) | 14 | 3 | 3 |
| *total* | 20362 | 3681 | 5394 |
### en-de
| Corpus | Train | Dev | Test |
|----------------------------------------------------------------------|--------|-------|-------|
| [Massive 1.1](https://huggingface.co/datasets/AmazonScience/massive) | 7536 | 1346 | 1955 |
### en-es
| Corpus | Train | Dev | Test |
|----------------------------------------------------------------------|--------|-------|-------|
| [Massive 1.1](https://huggingface.co/datasets/AmazonScience/massive) | 8415 | 1526 | 2202 |
### en-sv
| Corpus | Train | Dev | Test |
|----------------------------------------------------------------------|--------|-------|-------|
| [Massive 1.1](https://huggingface.co/datasets/AmazonScience/massive) | 7540 | 1360 | 1921 |
### en-fr
| Corpus | Train | Dev | Test |
|----------------------------------------------------------------------|--------|-------|-------|
| [Massive 1.1](https://huggingface.co/datasets/AmazonScience/massive) | 6800 | 1203 | 1757 |
### en-pt
| Corpus | Train | Dev | Test |
|----------------------------------------------------------------------|--------|-------|-------|
| [Massive 1.1](https://huggingface.co/datasets/AmazonScience/massive) | 7368 | 1296 | 1885 |
### en-hi
| Corpus | Train | Dev | Test |
|----------------------------------------------------------------------|--------|-------|-------|
| [Massive 1.1](https://huggingface.co/datasets/AmazonScience/massive) | 6702 | 1175 | 1747 |
### en-tr
| Corpus | Train | Dev | Test |
|----------------------------------------------------------------------|--------|-------|-------|
| [Massive 1.1](https://huggingface.co/datasets/AmazonScience/massive) | 8269 | 1474 | 2170 |
### en-ja
| Corpus | Train | Dev | Test |
|----------------------------------------------------------------------|--------|-------|-------|
| [Massive 1.1](https://huggingface.co/datasets/AmazonScience/massive) | 8066 | 1434 | 2085 |
### en-zh
| Corpus | Train | Dev | Test |
|----------------------------------------------------------------------|--------|-------|-------|
| [Massive 1.1](https://huggingface.co/datasets/AmazonScience/massive) | 8433 | 1513 | 2179 |
## Tools
Scripts used to generate this dataset can be found on [github](https://github.com/cartesinus/iva_mt).
## License
This is a composition of 7 datasets, and the license is as defined in original release:
- MASSIVE: [CC-BY 4.0](https://huggingface.co/datasets/AmazonScience/massive/blob/main/LICENSE)
- Leyzer: [CC BY-NC 4.0](https://github.com/cartesinus/leyzer/blob/master/LICENSE)
- OpenSubtitles: unknown
- KDE: [GNU Public License](https://l10n.kde.org/about.php)
- CCMatrix: no license given, therefore assuming it is LASER project license [BSD](https://github.com/facebookresearch/LASER/blob/main/LICENSE)
- Ubuntu: [GNU Public License](https://help.launchpad.net/Legal)
- Gnome: unknown
|
false | # Dataset Card for the ProofLang Corpus
## Dataset Summary
The ProofLang Corpus includes 3.7M proofs (558 million words) mechanically extracted from papers that were posted on [arXiv.org](https://arXiv.org) between 1992 and 2020.
The focus of this corpus is proofs, rather than the explanatory text that surrounds them, and more specifically on the *language* used in such proofs.
Specific mathematical content is filtered out, resulting in sentences such as `Let MATH be the restriction of MATH to MATH.`
This dataset reflects how people prefer to write (non-formalized) proofs, and is also amenable to statistical analyses and experiments with Natural Language Processing (NLP) techniques.
We hope it can serve as an aid in the development of language-based proof assistants and proof checkers for professional and educational purposes.
## Dataset Structure
There are multiple TSV versions of the data. Primarily, `proofs` divides up the data proof-by-proof, and `sentences` further divides up the same data sentence-by-sentence.
The `raw` dataset is a less-cleaned-up version of `proofs`. More usefully, the `tags` dataset gives arXiv subject tags for each paper ID found in the other data files.
* The data in `proofs` (and `raw`) consists of a `paper` ID (identifying where the proof was extracted from), and the `proof` as a string.
* The data in `sentences` consists of a `paper` ID, and the `sentence` as a string.
* The data in `tags` consists of a `paper` ID, and the arXiv subject tags for that paper as a single comma-separated string.
Further metadata about papers can be queried from arXiv.org using the paper ID.
In particular, each paper `<id>` in the dataset can be accessed online at the url `https://arxiv.org/abs/<id>`
## Dataset Size
* `proofs` is 3,094,779,182 bytes (unzipped) and has 3,681,893 examples.
* `sentences` is 3,545,309,822 bytes (unzipped) and has 38,899,132 examples.
* `tags` is 7,967,839 bytes (unzipped) and has 328,642 rows.
* `raw` is 3,178,997,379 bytes (unzipped) and has 3,681,903 examples.
## Dataset Statistics
* The average length of `sentences` is 14.1 words.
* The average length of `proofs` is 10.5 sentences.
## Dataset Usage
Data can be downloaded as (zipped) TSV files.
Accessing the data programmatically from Python is also possible using the `Datasets` library.
For example, to print the first 10 proofs:
```python
from datasets import load_dataset
dataset = load_dataset('proofcheck/prooflang', 'proofs', split='train', streaming='True')
for d in dataset.take(10):
print(d['paper'], d['proof'])
```
To look at individual sentences from the proofs,
```python
dataset = load_dataset('proofcheck/prooflang', 'proofs', split='train', streaming='True')
for d in dataset.take(10):
print(d['paper'], d['sentence'])
```
To get a comma-separated list of arXiv subject tags for each paper,
```python
from datasets import load_dataset
dataset = load_dataset('proofcheck/prooflang', 'tags', split='train', streaming='True')
for d in dataset.take(10):
print(d['paper'], d['tags'])
```
Finally, to look at a version of the proofs with less aggressive cleanup (straight from the LaTeX extraction),
```python
dataset = load_dataset('proofcheck/prooflang', 'raw', split='train', streaming='True')
for d in dataset.take(10):
print(d['paper'], d['proof'])
```
### Data Splits
There is currently no train/test split; all the data is in `train`.
## Dataset Creation
We started with the LaTeX source of 1.6M papers that were submitted to [arXiv.org](https://arXiv.org) between 1992 and April 2022.
The proofs were extracted using a Python script simulating parts of LaTeX (including defining and expanding macros).
It does no actual typesetting, throws away output not between `\begin{proof}...\end{proof}`, and skips math content. During extraction,
* Math-mode formulas (signalled by `$`, `\begin{equation}`, etc.) become `MATH`
* `\ref{...}` and variants (`autoref`, `\subref`, etc.) become `REF`
* `\cite{...}` and variants (`\Citet`, `\shortciteNP`, etc.) become `CITE`
* Words that appear to be proper names become `NAME`
* `\item` becomes `CASE:`
We then run a cleanup pass on the extracted proofs that includes
* Cleaning up common extraction errors (e.g., due to uninterpreted macros)
* Replacing more references by `REF`, e.g., `Theorem 2(a)` or `Postulate (*)`
* Replacing more citations with `CITE`, e.g., `Page 47 of CITE`
* Replacing more proof-case markers with `CASE:`, e.g., `Case (a).`
* Fixing a few common misspellings
## Additional Information
This dataset is released under the Creative Commons Attribution 4.0 licence.
Copyright for the actual proofs remains with the authors of the papers on [arXiv.org](https://arXiv.org), but these simplified snippets are fair use under US copyright law.
|
true |
# Dataset Card for Dataset Name
## Dataset Description
- **Homepage:** https://rucola-benchmark.com
- **Repository:** https://github.com/RussianNLP/RuCoLA
- **Paper:** https://aclanthology.org/2022.emnlp-main.348/
- **ArXiv:** https://arxiv.org/abs/2210.12814
- **Leaderboard:** https://rucola-benchmark.com/leaderboard
- **Point of Contact:** vmikhailovhse@gmail.com
- **Language:** Russian
### Dataset Summary
Russian Corpus of Linguistic Acceptability (RuCoLA) is a novel benchmark of 13.4k sentences labeled as acceptable or not. RuCoLA combines in-domain sentences manually collected from linguistic literature and out-of-domain sentences produced by nine machine translation and paraphrase generation models.
The motivation behind the out-of-domain set is to facilitate the practical use of acceptability judgments for improving language generation.
Each unacceptable sentence is additionally labeled with four standard and machine-specific coarse-grained categories: morphology, syntax, semantics, and hallucinations.
## Dataset Structure
### Supported Tasks and Leaderboards
- **Task:** binary classification.
- **Metrics:** MCC/Acc.
- **Leaderboard:** https://rucola-benchmark.com/leaderboard
### Languages
Russian.
### Data Instances
```
{
"id": 19,
"sentence": "Люк останавливает удачу от этого.",
"label": 0,
"error_type": "Hallucination",
"detailed_source": "WikiMatrix"}
}
```
The example in English for illustration purposes:
```
{
"id": 19,
"sentence": "Luck stops luck from doing this.",
"label": 0,
"error_type": "Hallucination",
"detailed_source": "WikiMatrix"}
}
```
### Data Fields
- ```id (int64)```: the sentence's id.
- ```sentence (str)```: the sentence.
- ```label (str)```: the target class. "1" refers to "acceptable", while "0" corresponds to "unacceptable".
- ```error_type (str)```: the coarse-grained violation category (Morphology, Syntax, Semantics, or Hallucination); "0" if the sentence is acceptable.
- ```detailed_source```: the data source.
### Data Splits
RuCoLA consists of the training, development, and private test sets organised under two subsets: in-domain (linguistic publications) and out-of-domain (texts produced by natural language generation models).
- ```train```: 7869 in-domain samples (```"data/in_domain_train.csv"```).
- ```validation```: 2787 in-domain and out-of-domain samples. The in-domain (```"data/in_domain_dev.csv"```) and out-of-domain (```"data/out_of_domain_dev.csv"```) validation sets are merged into ```"data/dev.csv"``` for convenience.
- ```test```: 2789 in-domain and out-of-domain samples (```"data/test.csv"```).
## Dataset Creation
### Curation Rationale
- **In-domain Subset:** The in-domain sentences and the corresponding authors’ acceptability judgments are *manually* drawn from fundamental linguistic textbooks, academic publications, and methodological materials.
- **Out-of-domain Subset:** The out-of-domain sentences are produced by nine open-source MT and paraphrase generation models.
### Source Data
<details>
<summary>Linguistic publications and resources</summary>
|Original source |Transliterated source |Source id |
|---|---|---|
|[Проект корпусного описания русской грамматики](http://rusgram.ru) | [Proekt korpusnogo opisaniya russkoj grammatiki](http://rusgram.ru/)|Rusgram |
|Тестелец, Я.Г., 2001. *Введение в общий синтаксис*. Федеральное государственное бюджетное образовательное учреждение высшего образования Российский государственный гуманитарный университет.|Yakov Testelets. 2001. Vvedeniye v obschiy sintaksis. Russian State University for the Humanities. |Testelets |
|Лютикова, Е.А., 2010. *К вопросу о категориальном статусе именных групп в русском языке*. Вестник Московского университета. Серия 9. Филология, (6), pp.36-76. |Ekaterina Lutikova. 2010. K voprosu o kategorial’nom statuse imennykh grup v russkom yazyke. Moscow University Philology Bulletin. |Lutikova |
|Митренина, О.В., Романова, Е.Е. and Слюсарь, Н.А., 2017. *Введение в генеративную грамматику*. Общество с ограниченной ответственностью "Книжный дом ЛИБРОКОМ". |Olga Mitrenina et al. 2017. Vvedeniye v generativnuyu grammatiku. Limited Liability Company “LIBROCOM”. |Mitrenina |
|Падучева, Е.В., 2004. *Динамические модели в семантике лексики*. М.: Языки славянской культуры.| Elena Paducheva. 2004. Dinamicheskiye modeli v semantike leksiki. Languages of Slavonic culture. |Paducheva2004 |
|Падучева, Е.В., 2010. *Семантические исследования: Семантика времени и вида в русском языке; Семантика нарратива*. М.: Языки славянской культуры. | Elena Paducheva. 2010. Semanticheskiye issledovaniya: Semantika vremeni i vida v russkom yazyke; Semantika narrativa. Languages of Slavonic culture.|Paducheva2010 |
|Падучева, Е.В., 2013. *Русское отрицательное предложение*. М.: Языки славянской культуры |Elena Paducheva. 2013. Russkoye otritsatel’noye predlozheniye. Languages of Slavonic culture. |Paducheva2013 |
|Селиверстова, О.Н., 2004. *Труды по семантике*. М.: Языки славянской культуры | Olga Seliverstova. 2004. Trudy po semantike. Languages of Slavonic culture.|Seliverstova |
| Набор данных ЕГЭ по русскому языку | Shavrina et al. 2020. [Humans Keep It One Hundred: an Overview of AI Journey](https://aclanthology.org/2020.lrec-1.277/) |USE5, USE7, USE8 |
</details>
<details>
<summary>Machine-generated sentences</summary>
<br>
**Datasets**
|Original source |Source id|
|---|---|
|Mikel Artetxe and Holger Schwenk. 2019. [Massively Multilingual Sentence Embeddings for Zero-Shot Cross-Lingual Transfer and Beyond](https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00288/43523/Massively-Multilingual-Sentence-Embeddings-for)|Tatoeba |
|Holger Schwenk et al. 2021. [WikiMatrix: Mining 135M Parallel Sentences in 1620 Language Pairs from Wikipedia](https://aclanthology.org/2021.eacl-main.115/)|WikiMatrix |
|Ye Qi et al. 2018. [When and Why Are Pre-Trained Word Embeddings Useful for Neural Machine Translation?](https://aclanthology.org/N18-2084/)|TED |
|Alexandra Antonova and Alexey Misyurev. 2011. [Building a Web-Based Parallel Corpus and Filtering Out Machine-Translated Text](https://aclanthology.org/W11-1218/)|YandexCorpus |
**Models**
[EasyNMT models](https://github.com/UKPLab/EasyNMT):
1. OPUS-MT. Jörg Tiedemann and Santhosh Thottingal. 2020. [OPUS-MT – Building open translation services for the World](https://aclanthology.org/2020.eamt-1.61/)
2. M-BART50. Yuqing Tang et al. 2020. [Multilingual Translation with Extensible Multilingual Pretraining and Finetuning](https://arxiv.org/abs/2008.00401)
3. M2M-100. Angela Fan et al. 2021. [Beyond English-Centric Multilingual Machine Translation](https://jmlr.org/papers/volume22/20-1307/20-1307.pdf)
[Paraphrase generation models](https://github.com/RussianNLP/russian_paraphrasers):
1. [ruGPT2-Large](https://huggingface.co/sberbank-ai/rugpt2large)
2. [ruT5](https://huggingface.co/cointegrated/rut5-base-paraphraser)
3. mT5. Linting Xue et al. 2021. [mT5: A Massively Multilingual Pre-trained Text-to-Text Transformer](https://aclanthology.org/2021.naacl-main.41/)
</details>
### Annotations
#### Annotation process
The out-of-domain sentences undergo a two-stage annotation procedure on [Toloka](https://toloka.ai), a crowd-sourcing platform for data labeling.
Each stage includes an unpaid training phase with explanations, control tasks for tracking annotation quality, and the main annotation task. Before starting, the worker is given detailed instructions describing the task, explaining the labels, and showing plenty of examples.
The instruction is available at any time during both the training and main annotation phases. To get access to the main phase, the worker should first complete the training phase by labeling more than 70% of its examples correctly. Each trained worker receives a page with five sentences, one of which is a control one.
We collect the majority vote labels via a dy- namic overlap from three to five workers after filtering them by response time and performance on control tasks.
- **Stage 1: Acceptability Judgments**
The first annotation stage defines whether a given sentence is acceptable or not. Access to the project is granted to workers certified as native speakers of Russian by Toloka and ranked top-60% workers according to the Toloka rating system.
Each worker answers 30 examples in the training phase. Each training example is accompanied by an explanation that appears in an incorrect answer.
The main annotation phase counts 3.6k machine-generated sentences. The pay rate is on average $2.55/hr, which is twice the amount of the hourly minimum wage in Russia. Each of 1.3k trained workers get paid, but we keep votes from only 960 workers whose annotation quality rate on the control sentences is more than 50%.
- **Stage 2: Violation Categories**
The second stage includes validation and annotation of sentences labeled unacceptable on Stage 1 according to five answer options: “Morphology”, “Syntax”, “Semantics”, “Hallucinations” and “Other”. The task is framed as a multi-label classification, i.e., the sentence may contain more than one violation in some rare cases or be re-labeled as acceptable.
We create a team of 30 annotators who are undergraduate BA and MA in philology and linguistics from several Russian universities. The students are asked to study the works on CoLA, TGEA, and hallucinations. We also hold an online seminar to discuss the works and clarify the task specifics. Each student undergoes platform-based training on 15 examples before moving onto the main phase of 1.3k sentences.
The students are paid on average $5.42/hr and are eligible to get credits for an academic course or an internship. This stage provides direct interaction between authors and students in a group chat. We keep submissions with more than 30 seconds of response time per page and collect the majority vote labels for each answer independently.
Sentences having more than one violation category or labeled as “Other” by the majority are filtered out.
### Personal and Sensitive Information
The annotators are warned about potentially sensitive topics in data (e.g., politics, culture, and religion).
## Considerations for Using the Data
### Social Impact of Dataset
RuCoLA may serve as training data for acceptability classifiers, which may benefit the quality of generated texts.
We recognize that such improvements in text generation may lead to misuse of LMs for malicious purposes. However, our corpus can be used to train adversarial defense and artificial text detection models.
We introduce a novel dataset for **research and development needs**, and the potential negative uses are not lost on us.
### Discussion of Biases
Although we aim to control the number of high-frequency tokens in the RuCoLA’s sentences, we assume that potential word frequency distribution shift between LMs’ pretraining corpora and our corpus can introduce bias in the evaluation.
Furthermore, linguistic publications represent a specific domain as the primary source of acceptability judgments. On the one hand, it can lead to a domain shift when using RuCoLA for practical purposes.
On the other hand, we observe moderate acceptability classification performance on the out-of-domain test, which spans multiple domains, ranging from subtitles to Wikipedia.
### Other Known Limitations
- **Data Collection**
Acceptability judgments datasets require a source of unacceptable sentences.
Collecting judgments from linguistic literature has become a standard practice replicated in multiple languages. However, this approach has several limitations. First, many studies raise concerns about the reliability and reproducibility of acceptability judgments. Second, the linguists’ judgments may limit data representativeness, as they may not reflect the errors that speakers tend to produce. Third, enriching acceptability judgments datasets is time-consuming, while creating new ones can be challenging due to limited resources, e.g., in low-resource languages.
- **Expert vs. Non-expert**
One of the open methodological questions on acceptability judgments is whether they should be collected from expert or non-expert speakers.
On the one hand, prior linguistic knowledge can introduce bias in reporting judgments. On the other hand, expertise may increase the quality of the linguists’ judgments over the ones of non-linguists. At the same time, the latter tend to be influenced by an individual’s exposure to ungrammatical language use.
The objective of involving students with a linguistic background is to maximize the annotation quality.
- **Fine-grained Annotation**
The coarse-grained annotation scheme of the RuCoLA’s unacceptable sentences relies on four major categories. While the annotation can be helpful for model error analysis, it limits the scope of LMs’ diagnostic evaluation concerning linguistic and machine-specific phenomena.
## Additional Information
### Dataset Curators
Correspondence: ```vmikhailovhse@gmail.com```
### Licensing Information
Our baseline code and acceptability labels are available under the Apache 2.0 license. The copyright (where applicable) of texts from the linguistic publications and resources remains with the original authors or publishers.
### Citation Information
```
@inproceedings{mikhailov-etal-2022-rucola,
title = "{R}u{C}o{LA}: {R}ussian Corpus of Linguistic Acceptability",
author = "Mikhailov, Vladislav and
Shamardina, Tatiana and
Ryabinin, Max and
Pestova, Alena and
Smurov, Ivan and
Artemova, Ekaterina",
booktitle = "Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing",
month = dec,
year = "2022",
address = "Abu Dhabi, United Arab Emirates",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2022.emnlp-main.348",
pages = "5207--5227",
abstract = "Linguistic acceptability (LA) attracts the attention of the research community due to its many uses, such as testing the grammatical knowledge of language models and filtering implausible texts with acceptability classifiers.However, the application scope of LA in languages other than English is limited due to the lack of high-quality resources.To this end, we introduce the Russian Corpus of Linguistic Acceptability (RuCoLA), built from the ground up under the well-established binary LA approach. RuCoLA consists of 9.8k in-domain sentences from linguistic publications and 3.6k out-of-domain sentences produced by generative models. The out-of-domain set is created to facilitate the practical use of acceptability for improving language generation.Our paper describes the data collection protocol and presents a fine-grained analysis of acceptability classification experiments with a range of baseline approaches.In particular, we demonstrate that the most widely used language models still fall behind humans by a large margin, especially when detecting morphological and semantic errors. We release RuCoLA, the code of experiments, and a public leaderboard to assess the linguistic competence of language models for Russian.",
}
```
### Other
Please refer to our [paper](https://aclanthology.org/2022.emnlp-main.348/) for more details. |
false |
# Dataset Card for construction-safety-gsnvb
** The original COCO dataset is stored at `dataset.tar.gz`**
## Dataset Description
- **Homepage:** https://universe.roboflow.com/object-detection/construction-safety-gsnvb
- **Point of Contact:** francesco.zuppichini@gmail.com
### Dataset Summary
construction-safety-gsnvb
### Supported Tasks and Leaderboards
- `object-detection`: The dataset can be used to train a model for Object Detection.
### Languages
English
## Dataset Structure
### Data Instances
A data point comprises an image and its object annotations.
```
{
'image_id': 15,
'image': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=640x640 at 0x2373B065C18>,
'width': 964043,
'height': 640,
'objects': {
'id': [114, 115, 116, 117],
'area': [3796, 1596, 152768, 81002],
'bbox': [
[302.0, 109.0, 73.0, 52.0],
[810.0, 100.0, 57.0, 28.0],
[160.0, 31.0, 248.0, 616.0],
[741.0, 68.0, 202.0, 401.0]
],
'category': [4, 4, 0, 0]
}
}
```
### Data Fields
- `image`: the image id
- `image`: `PIL.Image.Image` object containing the image. Note that when accessing the image column: `dataset[0]["image"]` the image file is automatically decoded. Decoding of a large number of image files might take a significant amount of time. Thus it is important to first query the sample index before the `"image"` column, *i.e.* `dataset[0]["image"]` should **always** be preferred over `dataset["image"][0]`
- `width`: the image width
- `height`: the image height
- `objects`: a dictionary containing bounding box metadata for the objects present on the image
- `id`: the annotation id
- `area`: the area of the bounding box
- `bbox`: the object's bounding box (in the [coco](https://albumentations.ai/docs/getting_started/bounding_boxes_augmentation/#coco) format)
- `category`: the object's category.
#### Who are the annotators?
Annotators are Roboflow users
## Additional Information
### Licensing Information
See original homepage https://universe.roboflow.com/object-detection/construction-safety-gsnvb
### Citation Information
```
@misc{ construction-safety-gsnvb,
title = { construction safety gsnvb Dataset },
type = { Open Source Dataset },
author = { Roboflow 100 },
howpublished = { \url{ https://universe.roboflow.com/object-detection/construction-safety-gsnvb } },
url = { https://universe.roboflow.com/object-detection/construction-safety-gsnvb },
journal = { Roboflow Universe },
publisher = { Roboflow },
year = { 2022 },
month = { nov },
note = { visited on 2023-03-29 },
}"
```
### Contributions
Thanks to [@mariosasko](https://github.com/mariosasko) for adding this dataset. |
false | # WikiDT: Wikipedia Table Document dataset for table extraction and visual question answering
## Dataset Description
- **Homepage:**
- **Repository:**
- **Paper:**
- **Leaderboard:**
- **Point of Contact:**
### Dataset Summary
The WikiDT contains multi-level annotations and labels for the question-answering task based on images. Meanwhile, as the questions are answered from some table on the image, and WikiDT provides the table annotation to facilitate the diagnosis of the models and decompose the problem, WikiDT can be also directly used as a
table recognition dataset.
The dataset contains 16,887 Wikipedia screenshot, which are segmented to 54,032 subpages since the full screenshots are potentially long. In total, there's 159,905 tables in the dataset. The number of question-answer samples is 70,652. Each QA sample contains triplets of <question, answer, full-page screenshot filename>, and is additionally annotated with retrieval labels (which subpage, and which table). 53,698 QA samples also have SQL annotation.
For each subpage, OCR and table extraction annotations from two sources are available. While rendering the screenshots, the ground truth table annotation is recorded. Meanwhile, to make the dataset realistic, we also requested OCR and table extraction from [Amazon Textract](https://aws.amazon.com/textract/) for each subpage (results obtained during Feb.28, 2023 - Mar.6, 2023).
### Languages
English
## Dataset Structure
Once downloaded, the WikiDT has the following parts. The downloaded files are around 77GB. Please ensure you have at least 160GB since we will be extract individual files from the tars.
```
.
├── WikiTableExtraction
│ ├── detection.partaa
│ ├── detection.partab
│ ├── detection.partac
│ ├── detection.partad
│ ├── detection.partae
│ ├── detection.partaf
│ ├── detection.partag
│ ├── structure.partaa
│ ├── structure.partab
│ ├── structure.partac
│ ├── structure.partad
│ └── structure.partae
├── images.partaa
├── images.partab
├── images.partac
├── images.partad
├── images.partae
├── images.partaf
├── images.partag
├── images.partah
├── images.partai
├── ocr.tar
├── samples
│ ├── test.json
│ ├── train.json
│ └── val.json
└── tsv.tar
```
Please concat the part files and extract them into respective folder. For example,
run
```
cd WikiTableExtraction/
cat detection.parta* | tar x
```
to extract the `detection` folder.
Once you extracted all the tar files, the WikiDT dataset has the following file structure.
```sh
+--WikiDT-dataset
| +--WikiTableExtraction
| | +--detection
| | | +--images # sub page images
| | | +--train # xml table bbox annotation
| | | +--test # xml table bbox annotation
| | | +--val # xml table bbox annotation
| | | images_filelist.txt # index of 54,032 images
| | | test_filelist.txt # index of 5,410 test samples
| | | train_filelist.txt # index of 43,248 train samples
| | | val_filelist.txt # index of 5,347 val samples
| | +--structure
| | | +--images # images cropped to table region
| | | +--train # xml table bbox annotation
| | | +--test # xml table bbox annotation
| | | +--val # xml table bbox annotation
| | | images_filelist.txt # index of 159,898 images
| | | test_filelist.txt # index of 15,989 test samples
| | | train_filelist.txt # index of 129,980 train samples
| | | val_filelist.txt # index of 15,991 val samples
| +--samples # in total 70,652 TableVQA samples from the three json files
| | +--train.json #
| | +--test.json #
| | +--val.json #
| +--images # full page image
| +--ocr # text and bbox for the table content
| | +--textract # detected by Amazon Textract API
| | +--web # extracted from HTML information
| +--tsv # extracted table in tsv format
| | +--textract # detected by Amazon Textract API
| | +--web # extracted from HTML information
```
### Table VQA annotation example
Here is an example of an xml table bbox annotation from `WikiDT-dataset/samples/[train|test|val].json/`.
```
{'all_ocr_files_textract': ['ocr/textract/16301437_page_seg_0.json',
'ocr/textract/16301437_page_seg_1.json'],
'all_ocr_files_web': ['ocr/web/16301437_page_seg_0.json',
'ocr/web/16301437_page_seg_1.json'],
'all_table_files_textract': ['tsv/textract/16301437_page_0.tsv',
'tsv/textract/16301437_page_1.tsv'],
'all_table_files_web': ['tsv/web/16301437_1.tsv', 'tsv/web/16301437_0.tsv'],
'answer': [['don johnson buckeye st. classic']],
'image': '16301437_page.png',
'ocr_retrieval_file_textract': 'ocr/textract/16301437_page_seg_0.json',
'ocr_retrieval_file_web': 'ocr/web/16301437_page_seg_0.json',
'question': 'Name the Event which has a Score of 209-197?',
'sample_id': '14190',
'sql_str': "SELECT `event` FROM cur_table WHERE `score` = '209-197' ",
'sub_page': ['16301437_page_seg_0.png', '16301437_page_seg_1.png'],
'sub_page_retrieved': '16301437_page_seg_0.png',
'subset': 'TFC',
'table_id': '2-16301437-1',
'table_retrieval_file_textract': 'tsv/textract/16301437_page_0.tsv',
'table_retrieval_file_web': 'tsv/web/16301437_1.tsv'}
```
### Table Detection annotation example
Here is an example of an xml table bbox annotation from `WikiDT-dataset/WikiTableExtraction/structure/[train|test|val]/`.
```xml
<annotation>
<folder />
<filename>204_147_page_crop_5.png</filename>
<source>WikiDT Dataset</source>
<size>
<width>788</width>
<height>540.0</height>
<depth>3</depth>
</size>
<object>
<name>table</name>
<rowspan />
<colspan />
<bndbox>
<xmin>10</xmin>
<ymin>10</ymin>
<xmax>778</xmax>
<ymax>530</ymax>
</bndbox>
</object>
<object>
<name>header row</name>
<rowspan />
<colspan />
<bndbox>
<xmin>10</xmin>
<ymin>10</ymin>
<xmax>778</xmax>
<ymax>33</ymax>
</bndbox>
</object>
<object>
<name>header cell</name>
<rowspan />
<colspan>10</colspan>
<bndbox>
<xmin>12</xmin>
<ymin>35</ymin>
<xmax>776</xmax>
<ymax>58</ymax>
</bndbox>
</object>
<object>
<name>table row</name>
<rowspan />
<colspan />
<bndbox>
<xmin>10</xmin>
<ymin>60</ymin>
<xmax>778</xmax>
<ymax>530</ymax>
</bndbox>
</object>
</annotation>
```
### Licensing Information
CC BY SA 3.0
### Contributors
[Hui Shi](mailto:hshi@ucsd.edu) (Work done during her internship at Amazon)
[Yusheng Xie](mailto:yushx@amazon.com) (corresponding person)
[Luis Goncalves](mailto:luisgonc@amazon.com)
|
false |
# Peanut Comic Strip Dataset (Snoopy & Co.)
This is a dataset Peanuts comic strips from `1950/10/02` to `2000/02/13`.
There are `77,456` panels extracted from `17,816` comic strips.
The dataset size is approximately `4.4G`.
Each row in the dataset contains the following fields:
- `image`: `PIL.Image` containing the extracted panel.
- `panel_name`: unique identifier for the row.
- `characters`: `tuple[str, ...]` of characters included in the comic strip the panel is part of.
- `themes`: `tuple[str, ...]` of theme in the comic strip the panel is part of.
- `color`: `str` indicating whether the panel is grayscale or in color.
- `caption`: [BLIP-2_FLAN-T5-XL](https://huggingface.co/docs/transformers/main/model_doc/blip-2) generated caption from the panel.
- `year`: `int` storing the year the specific panel was released.
> **FLAN-T5-XL has a commercial use license and so this dataset can be used for commercial projects. Alternatively use [this similar dataset](https://huggingface.co/datasets/afmck/peanuts-opt-6.7b) that uses OPT-6.7B as the caption pipeline's text model, however it does not permit commercial use.**
Character and theme information was extracted from [Peanuts Wiki (Fandom)](https://peanuts.fandom.com/wiki/Peanuts_Wiki) using [Beautiful Soup](https://www.crummy.com/software/BeautifulSoup/bs4/doc/).
Images were extracted from [Peanuts Search](https://peanuts-search.com/).
Only strips with the following characters were extracted:
```
- "Charlie Brown"
- "Sally Brown"
- "Joe Cool" # Snoopy alter-ego
- "Franklin"
- "Violet Gray"
- "Eudora"
- "Frieda"
- "Marcie"
- "Peppermint Patty"
- "Patty"
- "Pig-Pen"
- "Linus van Pelt"
- "Lucy van Pelt"
- "Rerun van Pelt"
- "Schroeder"
- "Snoopy"
- "Shermy"
- "Spike"
- "Woodstock"
- "the World War I Flying Ace" # Snoopy alter-ego
```
### Extraction Details
Panel detection and extraction was done using the following codeblock:
```python
def check_contour(cnt):
area = cv2.contourArea(cnt)
if area < 600:
return False
_, _, w, h = cv2.boundingRect(cnt)
if w / h < 1 / 2: return False
if w / h > 2 / 1: return False
return True
def get_panels_from_image(path):
panels = []
original_img = cv2.imread(path)
gray = cv2.cvtColor(original_img, cv2.COLOR_BGR2GRAY)
blur = cv2.GaussianBlur(gray, (5,5), 0)
thresh = cv2.threshold(blur, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)[1]
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3,3))
opening = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel, iterations=1)
invert = 255 - opening
cnts, _ = cv2.findContours(invert, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
idx = 0
for cnt in cnts:
if not check_contour(cnt): continue
idx += 1
x,y,w,h = cv2.boundingRect(cnt)
roi = original_img[y:y+h,x:x+w]
panels.append(roi)
return panels
```
`check_contour` will reject panels with `area < 600` or with aspect ratios larger than `2` or smaller than `0.5`.
Grayscale detection was done using the following codeblock:
```python
def is_grayscale(panel):
LAB_THRESHOLD = 10.
img = cv2.cvtColor(panel, cv2.COLOR_RGB2LAB)
_, ea, eb = cv2.split(img)
de = abs(ea - eb)
mean_e = np.mean(de)
return mean_e < LAB_THRESHOLD
```
Captioning was done using the standard BLIP-2 pipeline shown in the [Huggingface docs](https://huggingface.co/docs/transformers/main/model_doc/blip-2) using beam search over 10 beams and a repetition penalty of `2.0`.
Raw captions are extracted and no postprocessing is applied. You may wish to normalise captions (such as replacing "cartoon" with "peanuts cartoon") or incorporate extra metadata into prompts. |
false | # Dataset Card for "LogicInference_OA"
This is an re-produce of the dataset from LogicInference Dataset in paper: https://openreview.net/pdf?id=HAGeIS_Lcg9.
The github page of LogicInference Dataset: https://github.com/google-research/google-research/tree/master/logic_inference_dataset.
This dataset is aimed to offer more dataset for Open Assistant project, depending on their demands, there three columns: INSTRUCTION, RESPONSE, SOURCE.
The results in this dataset is a little different from which was introduced in the original paper:
1.For all three splits (IID/OOD/length), only IID is used. In the original paper, it seems that model can reach better performance with data generated by this split method.
2.In the original paper, there are two form of responses: LOGICINFERENCE<sub>b</sub> (with the answer at the beginning) and LOGICINFERENCE<sub>e</sub> (with the answer at the end). This dataset uses LOGICINFERENCE<sub>e</sub>, that means: for all questions, the model will first do logic inference, and give the final answer at the end.
3.The original paper, some parameters in generate_dataset.py are:
N_INFERENCE_PROBLEMS = 5000
N_VARIATIONS = 25
N_EXAMPLES = 200000
TRAIN_RATIO = 0.9
LENGTH_SPLIT_THRESHOLD = 4
RANDOM_SEED = 0
I choose some new parameters:
N_INFERENCE_PROBLEMS = 10000
N_VARIATIONS = 25
N_EXAMPLES = 55000
TRAIN_RATIO = 1
LENGTH_SPLIT_THRESHOLD = 4
RANDOM_SEED = 1111
The original script generated 4814 different inference problems and extended all those inference problems to around 200,000 Q-A pairs. My settings generated 5491 different inference problems and extended them to around 54,607 Instruction-Response pairs. I think for Open Assistant projects, maybe the number of different inference problems is more important, and generated many similar Instruction-Response pairs will only add training time and doesn't make much sense. |
false | A collection of images from a very silly cat, these are all from @fatfatmillycat in twitter. Intended to be used with stable-diffusion-v1-4 |
false |
# Description
Dataset used to train our mT5 based model for machine translation, extracted from various text sources of National Library of Azerbaijan: [mT5-translation-enaz](https://huggingface.co/learningmachineaz/mt5-enaz-10m) \
It has only clean texts. Wiki articles wasn't used as they contain a lot of irrelevant data.
| Key point | Info |
|-------------------------|---------|
| Rows | ~10mil. EN-AZ sentence pairs |
| Size | 975M (zipped) / 2.8G (unzipped) |
| Format | TSV (tab separated pairs) |
| English | Google Translate |
| Azerbaijani | Original cleaned text |
## Author
Collected and prepared by [Renat Kalimulin](https://www.linkedin.com/in/rinat-kalimulin-16853358/)
|
false |
# Dataset Card for "podcasts-ner-es"
## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
- [Team members](#team-members)
## Dataset Description
- **Homepage:**
- **Repository:**
- **Paper:**
- **Leaderboard:**
- **Point of Contact:**
### Dataset Summary
This dataset comprises of small text snippets extracted from the "Deforme Semanal" podcast,
accompanied by annotations that identify the presence of a predetermined set of entities.
The purpose of this dataset is to facilitate Named Entity Recognition (NER) tasks.
The dataset was created to aid in the identification of entities such as famous people, books, or films in podcasts.
The transcription of the audio was first done, followed by annotation with GPT-3 and curation with Argilla.
The dataset is in Spanish, covering mostly topics such as love, feminism, and art, which are the main themes of the podcast.
### Supported Tasks and Leaderboards
Named Entity Recognition
### Languages
The dataset is in Spanish and the language used is primarily informal.
It is important to note that the language may include aggressive or offensive content.
## Dataset Structure
### Data Instances
```
{
"text":"Tengo 39 años, pues, ya veré cuándo yo quiero dejar de comer ternera, está mal, porque hay sobre explotación y todo esto, muy mal."
"annotation": [ { "end": 13, "label": "DATES", "start": 6 } ]
"id": "53c4748e-dbd2-4cf5-946f-d134b0bf6155"
}
```
### Data Fields
`text`: Snippet of text of no more than 512 characters extracted from a podcast episode.
`id`: Unique identification number for each instance in the dataset.
`annotation`: List of dictonary-like format with the following fields:
- `end`: end character of the entity ocurrence in the text.
- `start`: start character of the entity ocurrence in the text.
- `label`: label for the entity from the predefined set of entities. The label of the entities is one of:
'people', 'products', 'books', 'animals', 'organizations', 'topics', 'dates', 'places', 'artista', 'objects','songs', and 'films'.
### Data Splits
The dataset was shuffled and split using the `train_test_split` function from the Hugging Face datasets library.
The split was made with a train size of 0.8 and a seed of 42.
## Dataset Creation
### Curation Rationale
We created this dataset with the aim of making the information from our favorite podcasts more accessible, as retrieving information from audio formats can be challenging.
We chose to focus on the Named Entity Recognition (NER) task as it was relatively easy to annotate and validate.
### Source Data
#### Initial Data Collection and Normalization
We collected the data from a playlist on YouTube containing approximately 15 episodes of the "Deforme Semanal" podcast.
You can find the playlist at this [link](https://www.youtube.com/playlist?list=PLLbN7SMQhMVZoXhtQ00AyebQE_-ttDrs9).
We then transcribed the audio stream using OpenAI's Whisper (medium size) and split the resulting text files
into chunks of less than 512 characters.
### Annotations
#### Annotation process
To annotate the texts, we used OpenAI's API and GPT-3, with the following prompt:
```
Perform named entity recognition in Spanish. The classes are books, films, video games, songs, places, dates, topics, organizations, and people. The output should follow the format:
[{'class': 'people', 'text': 'name of the person'}, {'class': 'books', 'start': 'name of the book'}]
Sentence:
```
Finally, to ensure the quality of the dataset, we validated the annotations using Argilla by checking that the tokens were classified
correctly.
## Considerations for Using the Data
### Discussion of Biases
The dataset was obtained from the "Deforme Semanal" podcast, which primarily focuses on art, feminism, and culture.
As a result, the data is directly related to the topics and individuals discussed in these contexts. Additionally,
the language used in the podcast is informal and can be aggressive or offensive at times, which may be reflected in the dataset.
Although we attempted to minimize these biases during the validation process, their effectiveness is likely limited.
### Other Known Limitations
One issue that we have encountered with the token/entity data is that there can be some ambiguity in terms of their distinctions.
In some cases, it may not be clear how to differentiate between two tokens or entities, which can impact the accuracy
and effectiveness of models trained on this data.
Furthermore, the dataset size is relatively small, which can pose a challenge when it comes to training machine learning models.
With a limited amount of data, it can be difficult to capture the full range of variations and patterns in the data,
and overfitting can become a concern. This is especially true when dealing with complex models that require a large
amount of data to train effectively.
## Team members
[David Mora](https://huggingface.co/DavidFM43)
[Sergio Perez](https://huggingface.co/sergiopperez)
[Albeto Fernandez](https://huggingface.co/AlbertoFH98)
|
false |
# Dataset Card for Online Shoppers Purchasing Intention Dataset
## Dataset Description
- **Homepage**: https://archive-beta.ics.uci.edu/dataset/468/online+shoppers+purchasing+intention+dataset
### Dataset Summary
This dataset is a reupload of the Online Shoppers Purchasing Intention Dataset from the [UCI Machine Learning Repository](https://archive-beta.ics.uci.edu/).
> **NOTE:** The information below is from the original dataset description from UCI's website.
>
> ### Overview
>
> Of the 12,330 sessions in the dataset, 84.5% (10,422) were negative class samples that did not end with shopping,
> and the rest (1908) were positive class samples ending with shopping.
>
> #### Additional Information
>
> The dataset consists of feature vectors belonging to 12,330 sessions. The dataset was formed so that
> each session would belong to a different user in a 1-year period to avoid any tendency to a specific campaign,
> special day, user profile, or period.
|
true |
```bib
@inproceedings{lamprinidis2021universal,
title={Universal Joy A Dataset and Results for Classifying Emotions Across Languages},
author={Lamprinidis, Sotiris and Bianchi, Federico and Hardt, Daniel and Hovy, Dirk},
year={2021},
volume={11th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis (WASSA 2021)}
organization={Association for Computational Linguistics}
}
``` |
false |
# Dataset Card for Dating historical color images
## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Homepage:** http://graphics.cs.cmu.edu/projects/historicalColor/
- **Repository:**
- **Paper:** https://doi.org/10.1007/978-3-642-33783-3_36
- **Leaderboard:**
- **Point of Contact:**
### Dataset Summary
> We introduce the task of automatically estimating the age of historical color photographs. We suggest features which attempt to capture temporally discriminative information based on the evolution of color imaging processes over time and evaluate the performance of both these novel features and existing features commonly utilized in other problem domains on a novel historical image data set. For the challenging classification task of sorting historical color images into the decade during which they were photographed, we demonstrate significantly greater accuracy than that shown by untrained humans on the same data set. Additionally, we apply the concept of data-driven camera response function estimation to historical color imagery, demonstrating its relevance to both the age estimation task and the popular application of imitating the appearance of vintage color photography.
### Supported Tasks and Leaderboards
This dataset is intended to train image classification or regression models to predict the time period in which color photographs were taken. This task could be approached either as a classification task or could be approached as an image regression task.
## Dataset Structure
### Data Instances
[More Information Needed]
### Data Fields
[More Information Needed]
### Data Splits
There is a single training split since the original dataset doesn't define a train-test split.
## Dataset Creation
### Curation Rationale
[More Information Needed]
### Source Data
#### Initial Data Collection and Normalization
> Beginning with a collection of approximately 230,000 Flickr images taken prior to 1980, we perform automated removal of monochromatic images. The remaining images are manually inspected to remove non-photographic content (e.g. scans of vintage artwork) and any remaining monochromatic images. Finally, a random subsampling and decimation is performed to create a dataset containing an equal number of historical color images for each decade (1,375 images total).
### Annotations
#### Annotation process
Annotations are based on metadata available on Flickr.
#### Who are the annotators?
It appears that annotations are sourced via Flickr, so the annotators are presumed to be those uploading the images to Flickr. This will include individuals as well as cultural heritage institutions.
### Personal and Sensitive Information
[More Information Needed]
## Considerations for Using the Data
### Social Impact of Dataset
[More Information Needed]
### Discussion of Biases
[More Information Needed]
### Other Known Limitations
[More Information Needed]
## Additional Information
### Dataset Curators
[More Information Needed]
### Licensing Information
[More Information Needed]
### Citation Information
[More Information Needed]
### Contributions
Thanks to [@davanstrien](https://github.com/davanstrien) for adding this dataset. |
false | # AutoTrain Dataset for project: aymara-t5-small
## Dataset Description
This dataset has been automatically processed by AutoTrain for project aymara-t5-small.
### Languages
The BCP-47 code for the dataset's language is unk.
## Dataset Structure
### Data Instances
A sample from this dataset looks as follows:
```json
[
{
"feat_Lang": "Spanish",
"feat_langcode": "es",
"feat_Source": "Janiw sartasipk\u00e4ti aka mayiw phuqasi\u00f1apkama, presidentex nanakamp tantachaspan ukhamarak tama irnaqir jaqinakar tantachpan, kunawsas ukat kunjamraks munat wila masinakasar qallantani thaxta\u00f1xa sasaw \u201d huelga lurir Margarita L\u00f3pez mamax arsuwayat\u00e4na.",
"target": "\u201cNo nos iremos hasta que nuestros casos se hayan resuelto, que el presidente se re\u00fana con nosotros y que re\u00fana a un grupo de tragbajo para decirnos cu\u00e1ndo y c\u00f3mo empezar\u00e1n a encontrar a nuestros seres queridos \u201d, declar\u00f3 la huelguista de hambre Margarita L\u00f3pez.",
"source": "translate Aymara to Spanish: Erwin C blog Latino Americano uka tuqinkiriw m\u00e4 huelga lurir mamaru jawsayawayi:"
},
{
"feat_Lang": "English",
"feat_langcode": "en",
"feat_Source": "Credit: Heidi Shin.",
"target": "Cr\u00e9dito: Heidi Shun.",
"source": "translate English to Aymara: Credit: Heidi Shin."
}
]
```
### Dataset Fields
The dataset has the following fields (also called "features"):
```json
{
"feat_Lang": "Value(dtype='string', id=None)",
"feat_langcode": "Value(dtype='string', id=None)",
"feat_Source": "Value(dtype='string', id=None)",
"target": "Value(dtype='string', id=None)",
"source": "Value(dtype='string', id=None)"
}
```
### Dataset Splits
This dataset is split into a train and validation split. The split sizes are as follow:
| Split name | Num samples |
| ------------ | ------------------- |
| train | 28121 |
| valid | 7031 |
|
true |
# Dataset Card for turkish-nlp-suite/beyazperde-top-300-movie-reviews
<img src="https://raw.githubusercontent.com/turkish-nlp-suite/.github/main/profile/beyazPerde.png" width="20%" height="20%">
## Dataset Description
- **Repository:** [BeyazPerde Top 300 Movie Reviews](https://github.com/turkish-nlp-suite/BeyazPerde-Movie-Reviews/)
- **Paper:** [ACL link]()
- **Dataset:** BeyazPerde Top 300 Movie Reviews
- **Domain:** Social Media
### Dataset Summary
Beyazperde Movie Reviews offers Turkish sentiment analysis datasets that is scraped from popular movie reviews website Beyazperde.com. Top 300 Movies include audience reviews about best 300 movies of all the time. Here's the star rating distribution:
| star rating | count |
|---|---|
| 0.5 | 1.657 |
| 1.0 | 535 |
| 1.5 | 273 |
| 2.0 | 608 |
| 2.5 | 2.439 |
| 3.0 |2.277 |
| 3.5 | 5.550 |
| 4.0 | 13.248 |
| 4.5 | 10.077 |
| 5.0 | 17.351 |
| total | 54.015 |
As one sees, this dataset is highly unbalanced, number of 4 and 5 star ratings are much higher than 0, 1, 2 and 3 star reviews. This dataset offers the challenge of understanding the sentiment in a refined way, dissecting the positive sentiment into "very positive" or "okayish positive".
### Dataset Instances
An instance of this dataset looks as follows:
```
{
"movie": "Bay Evet",
"text": "Tam kıvamında çok keyifli bir film",
"rating": 4
}
```
### Data Split
| name |train|validation|test|
|---------|----:|---:|---:|
|BeyazPerde Top 300 Movie Reviews|44015|5000|5000|
### Citation
Coming soon
|
false | # Dataset Card for "code-search-net-javascript"
## Dataset Description
- **Homepage:** None
- **Repository:** https://huggingface.co/datasets/Nan-Do/code-search-net-JavaScript
- **Paper:** None
- **Leaderboard:** None
- **Point of Contact:** [@Nan-Do](https://github.com/Nan-Do)
### Dataset Summary
This dataset is the JavaScript portion of the CodeSarchNet annotated with a summary column.
The code-search-net dataset includes open source functions that include comments found at GitHub.
The summary is a short description of what the function does.
### Languages
The dataset's comments are in English and the functions are coded in JavaScript
### Data Splits
Train, test, validation labels are included in the dataset as a column.
## Dataset Creation
May of 2023
### Curation Rationale
This dataset can be used to generate instructional (or many other interesting) datasets that are useful to train LLMs
### Source Data
The CodeSearchNet dataset can be found at https://www.kaggle.com/datasets/omduggineni/codesearchnet
### Annotations
This datasets include a summary column including a short description of the function.
#### Annotation process
The annotation procedure was done using [Salesforce](https://huggingface.co/Salesforce) T5 summarization models.
A sample notebook of the process can be found at https://github.com/Nan-Do/OpenAssistantInstructionResponsePython
The annontations have been cleaned to make sure there are no repetitions and/or meaningless summaries. (some may still be present in the dataset)
### Licensing Information
Apache 2.0 |
false | # Dataset Card for xOA22 - Multilingual Prompts from OpenAssistant
### Dataset Summary
xOA22 consists of 22 prompts originally shown in Appendix E, page 25 of the [OpenAssistant Conversations paper](https://arxiv.org/pdf/2304.07327.pdf). These 22 prompts were then manually translated by volunteers into 5 languages: Arabic, Simplified Chinese, French, Hindi and Spanish.
These prompts were originally created for human evaluations of the multilingual abilities of [BLOOMChat](https://huggingface.co/sambanovasystems/BLOOMChat-176B-v1). Since not all prompts could be directly translatable due to cultural and linguistic differences, volunteers were encouraged to make appropriate substitutions and modifications that would maintain the intent of the original English prompt. As this was largely a collaborative, volunteer-led effort, this led to some discrepancies in the number of prompts per language. We make note of major departures from the original English prompt below.
### Languages
- Arabic (ar)
- English (en)
- Spanish (es)
- French (fr)
- Hindi (hi)
- Chinese (zh)
## Dataset Structure
### Data Fields
- `prompt`: manually translated prompt text. The English split is un-modified from the OpenAssistant Converstaions paper.
### Data Splits
The xOA22 dataset has 6 splits, one for each language. Below are the statistics for each split
| Dataset Split | Number of Instances in Split |
| ------------- | ---------------------------- |
| ar | 22 |
| en | 22 |
| es | 22 |
| fr | 22 |
| hi | 24 |
| zh | 24 |
### Translation Notes
Below are notes from volunteer translators. Note that the Hindi split does not include Prompt 4 in the English split.
- Arabic
- Prompt 12: Second part of the sentence was translated to “Please mainly limit the games to ones that can be played on a PC at home”, discarding the mentions of emulation.
- Prompt 19: Not sure how to translate "navigation system" to Arabic, I used Google Translate for this one.
- Spanish
- Prompt 12: IDK how to say crafting system in Spanish. I've always said crafteo.
- Prompt 21: Not sure how to translate niche, went for "developing a topic" instead
- Prompt 22: Hacking - hackeo? It's what I'd say colloquially in Spanish, but not sure if it's the right thing to use here
- French
- No translation notes
- Hindi
- Prompt 1: Replace "GLaDOS" with "Ravan", a famous antagonist from mythology
- Prompt 4: This prompt was left untranslated, and so **is missing from the Hindi split**. Translator reasons are as follows: generally people won't ask this in Hindi. Code writing community is generally English aware and they are most probably going to ask this question in English.
- Prompt 9: Corresponds to English Prompt 10. Specific names changed to well known persons in Hindi speaking world.
- Prompt 11: Corresponds to English Prompt 12. I removed "in depth crafting system", "directly or through emulation"
- Prompt 21: Corresponds to English Prompt 22. I removed "social security numbers", "Google", and "Apple"
- Prompt 22: This is a Hindi-specific prompt. The English translation is: "write me a poem on monsoon in 100 words"
- Prompt 23: This is a Hindi-specific prompt. The English translation is: "write me a recipe for butter chicken"
- Prompt 24: This is a Hindi-specific prompt. The English translation is: "How do I go from Delhi to Jaipur? Bus or car? Details please."
- Chinese
- Prompt 1: Changed GLaDOS to a fictional species from the Chinese sci-fi series The Three Body Problem
- Prompt 3: Didn't specify whether it's an imaginary world or real world to make it more natural in Chinese. The meaning is basically the same.
- Prompt 5: Animal rennet -> 人工奶酪 as the former is not familiar to most people
- Prompt 9: Translated "king" as "emperor" to align with Chinese history
- Prompt 10: Joe Biden & Joe Rogan -> 毛晓彤 & 光晓彤
- Prompt 11: Shakespeare -> 鲁迅
- Prompt 12: "sci-fi ships" -> starship (巨型星际飞船)
- Prompt 21: YouTube -> b站
- Prompt 22: social security number -> 身份证
- Prompt 23: This is a Chinese-specific prompt. The English translation is: "Explain Kubernetes in simple terms. Explain to me like I'm 11 years old."
- Prompt 24: This is a Chinese-specific prompt. The English translation is: "I will provide you with an argument or opinion of mine. I want you to criticize it as if you were Elon Musk".
- Translator note: I don't think there is a good counterpart entrepreneur like Elon Musk in China. Jack Ma is as wealthy and powerful as Elon Musk but they have quite different perspectives. So instead of finding an actual counterpart in China, we need to understand the characteristics of Elon Musk and translate accordingly.
### Curation Rationale
These prompts were originally curated in order to test the multilingual abilities of the BLOOMChat model. The model's responses to these translated prompts were compared to responses from other open-source chat models in a human evaluation study. Therefore, emphasis was placed on making translations as natural and understandable as possible to native speakers in order to emulate a chat setting, and we accepted feedback and modifications to the prompts from our volunteers.
### Dataset Curators
### Contributions
### Source Data
Appendix E, page 25 of ["OpenAssistant Conversations - Democratizing Large Language Model Alignment"](https://arxiv.org/pdf/2304.07327.pdf) |
true | |
false | # Dataset Card for "yelp_review_full"
[More Information needed](https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards) |
false |
## FinQA 한국어 번역본
Question, Answer 총 6252 Rows |
false |
# midjourney-v5-202304-clean
## 简介 Brief Introduction
非官方的，对Kaggle (Midjourney User Prompts & Generated Images (250k))[https://www.kaggle.com/datasets/succinctlyai/midjourney-texttoimage?select=general-01_2022_06_20.json] 上的数据集进行了清理，一共有 248,167对。
Unofficially, a cleanup of the dataset on Kaggle (Midjourney User Prompts & Generated Images (250k))[https://www.kaggle.com/datasets/succinctlyai/midjourney-texttoimage?select=general-01_2022_06_20.json] yielded 248,167 pairs.
## 数据集信息 Dataset Information
我做了一些清洗，清理出了两个文件：
- ori.parquet （145,918对，midjourney的四格图）
- upscaled.parquet （102,249对，使用了高清指令的图，这意味着这个图更受欢迎。）
I did some cleaning and cleaned out two files:
- ori_prompts_df.parquet (145,918 pairs, midjourney's four-frame diagrams)
- upscaled_prompts_df.parquet (102,249 pairs, graphs that use the Upscale command, which means this one is more popular.)
## 列信息 Column Information
1. `content` (内容): 这一列包含了消息的主要内容，可能包括文本、链接、或者其他元素。
This column contains the main content of the message, which may include text, links, or other elements.
2. `url` (网址): 这一列包含了附件的URL，通常是图片或者其他文件。
This column contains the URL of the attachment, usually an image or other file.
3. `proxy_url` (代理网址): 这一列包含了附件的代理URL，这个URL可以用来访问附件，即使在原始URL不可用的情况下。
This column contains the proxy URL of the attachment, which can be used to access the attachment even when the original URL is not available.
4. `width` (宽度): 这一列包含了附件的宽度，通常是图片的宽度。
This column contains the width of the attachment, usually the width of an image.
5. `height` (高度): 这一列包含了附件的高度，通常是图片的高度。
This column contains the height of the attachment, usually the height of an image.
6. `date` (日期): 这一列包含了消息的发送日期和时间。
This column contains the date and time the message was sent.
7. `message_type` (消息类型): 这一列包含了消息的类型，例如是否是初始消息、变体请求或者是放大请求。
This column contains the type of the message, such as whether it is an initial message, a variation request, or an upscale request.
8. `content_links` (内容链接): 这一列包含了消息内容中的所有链接。
This column contains all the links in the message content.
9. `prompt` (提示): 这一列包含了消息中的主要提示，通常是用户输入的文本。
This column contains the main prompt in the message, usually the text input by the user.
10. `prompt_additions` (提示补充): 这一列包含了消息中的提示补充，这些补充可能包括额外的信息或者指示。
This column contains the prompt additions in the message, these additions may include extra information or instructions.
11. `user_name` (用户名): 这一列包含了发送消息的用户的用户名。
This column contains the username of the user who sent the message.
12. `aspect` (宽高比): 这一列包含了附件的宽高比，通常是图片的宽高比。
This column contains the aspect ratio of the attachment, usually the aspect ratio of an image.
13. `clean_prompts` (清理后的提示): 这一列包含了清理后的提示，其中已经删除了所有的链接和奇怪的字符。
This column contains the cleaned prompts, where all links and weird characters have been removed.
|
false | |
false |
# stacked-summaries/onlystacked-xsum-1024
Same thing as [stacked-summaries/stacked-xsum-1024](https://huggingface.co/datasets/stacked-summaries/stacked-xsum-1024) but filtered such that `is_stacked=True`. Please refer to the original dataset for info and to raise issues if needed.
Basic info on `train` split:
```
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 116994 entries, 0 to 116993
Data columns (total 6 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 document 116994 non-null string
1 summary 116994 non-null string
2 id 116994 non-null Int64
3 chapter_length 116994 non-null Int64
4 summary_length 116994 non-null Int64
5 is_stacked 116994 non-null boolean
dtypes: Int64(3), boolean(1), string(2)
memory usage: 5.0 MB
```
_If you are interested in subscribing to this dataset on OnlyStacks, please reach out._ |
false |
# To join all training set files together
run `python join_dataset.py` file, final result will be `join_dataset.json` file |
false |
### Dataset Summary
Korean translation of Guanaco via the DeepL API
Note: There are cases where multilingual data has been converted to monolingual data during batch translation to Korean using the API.
Below is Guanaco's README.
----
This dataset is a subset of the Open Assistant dataset, which you can find here: https://huggingface.co/datasets/OpenAssistant/oasst1/tree/main
This subset of the data only contains the highest-rated paths in the conversation tree, with a total of 9,846 samples.
This dataset was used to train Guanaco with QLoRA.
For further information, please see the original dataset.
License: Apache 2.0 |
true | # OpenBookQA
- Source: https://huggingface.co/datasets/openbookqa
- Num examples:
- 4,957 (train)
- 500 (validation)
- 500 (test)
- Language: English
```python
from datasets import load_dataset
load_dataset("vietgpt/piqa_en")
```
- Format for GPT-3
```python
def preprocess_gpt3(sample):
question_stem = sample['question_stem']
choices = sample['choices']['text']
answerKey = sample['answerKey']
if answerKey == 'A':
output = f'\n<|correct|> {choices[0]}\n<|incorrect|> {choices[1]}\n<|incorrect|> {choices[2]}\n<|incorrect|> {choices[3]}'
elif answerKey == 'B':
output = f'\n<|correct|> {choices[1]}\n<|incorrect|> {choices[0]}\n<|incorrect|> {choices[2]}\n<|incorrect|> {choices[3]}'
elif answerKey == 'C':
output = f'\n<|correct|> {choices[2]}\n<|incorrect|> {choices[0]}\n<|incorrect|> {choices[1]}\n<|incorrect|> {choices[3]}'
else:
output = f'\n<|correct|> {choices[3]}\n<|incorrect|> {choices[0]}\n<|incorrect|> {choices[1]}\n<|incorrect|> {choices[2]}'
return {'text': f'<|startoftext|><|context|> {question_stem} <|answer|> {output} <|endoftext|>'}
"""
<|startoftext|><|context|> The sun is responsible for <|answer|>
<|correct|> plants sprouting, blooming and wilting
<|incorrect|> puppies learning new tricks
<|incorrect|> children growing up and getting old
<|incorrect|> flowers wilting in a vase <|endoftext|>
"""
``` |
false | |
false | # Dataset Card for Dataset Name
## Dataset Description
- **Homepage:**
- **Repository:**
- **Paper:**
- **Leaderboard:**
- **Point of Contact:**
### Dataset Summary
This dataset card aims to be a base template for new datasets. It has been generated using [this raw template](https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/templates/datasetcard_template.md?plain=1).
### Supported Tasks and Leaderboards
[More Information Needed]
### Languages
[More Information Needed]
## Dataset Structure
### Data Instances
[More Information Needed]
### Data Fields
[More Information Needed]
### Data Splits
[More Information Needed]
## Dataset Creation
### Curation Rationale
[More Information Needed]
### Source Data
#### Initial Data Collection and Normalization
[More Information Needed]
#### Who are the source language producers?
[More Information Needed]
### Annotations
#### Annotation process
[More Information Needed]
#### Who are the annotators?
[More Information Needed]
### Personal and Sensitive Information
[More Information Needed]
## Considerations for Using the Data
### Social Impact of Dataset
[More Information Needed]
### Discussion of Biases
[More Information Needed]
### Other Known Limitations
[More Information Needed]
## Additional Information
### Dataset Curators
[More Information Needed]
### Licensing Information
[More Information Needed]
### Citation Information
[More Information Needed]
### Contributions
[More Information Needed] |
false |
# Reasonix
Reasonix is a collection of 11,000 pairs of "Question Answer" and "Choice with Answer" gathered from different datasets. |
false |
# Dataset Card for HEADLINES
## Dataset Description
- **Homepage:** [Dell Research homepage](https://dell-research-harvard.github.io/)
- **Repository:** [Github repository](https://github.com/dell-research-harvard)
- **Paper:** [arxiv submission](https://arxiv.org/abs/tbd)
- **Point of Contact:** [Melissa Dell](mailto:melissadell@fas.harvard.edu)
#### Dataset Summary
HEADLINES is a massive English-language semantic similarity dataset, containing 396,001,930 pairs of different headlines for the same newspaper article, taken from historical U.S. newspapers, covering the period 1920-1989.
#### Languages
The text in the dataset is in English.
## Dataset Structure
Each year in the dataset is divided into a distinct file (eg. 1952_headlines.json), giving a total of 70 files.
The data is presented in the form of clusters, rather than pairs to eliminate duplication of text data and minimise the storage size of the datasets. Below we give an example of how to convert the dataset into pairs.
#### Dataset Instances
An example from the HEADLINES dataset looks like:
```python
{
"headline": "FRENCH AND BRITISH BATTLESHIPS IN MEXICAN WATERS",
"group_id": 4
"date": "May-14-1920",
"state": "kansas",
}
```
#### Dataset Fields
- `headline`: headline text.
- `date`: the date of publication of the newspaper article, as a string in the form mmm-DD-YYYY.
- `state`: state of the newspaper that published the headline.
- `group_id`: a number that is shared with all other headlines for the same article. This number is unique across all year files.
## Usage
The whole dataset can be easily downloaded using the `datasets` library.
```
from datasets import load_dataset
dataset_dict = load_dataset('dell-research-harvard/headlines-semantic-similarity')
```
If you just want to load specific files, you can specify these in the command.
```
from datasets import load_dataset
load_dataset(
'dell-research-harvard/headlines-semantic-similarity',
data_files=["1929_headlines.json", "1989_headlines.json"]
)
```
## Dataset Creation
### Source Data
The dataset was constructed using a large corpus of newly digitized articles from off-copyright, local U.S. newspapers.
Many of these newspapers reprint articles from newswires, such as the Associated Press, but the headlines are written locally.
The dataset comprises different headlines for the same article.
#### Initial Data Collection and Normalization
To construct HEADLINES, we digitize front pages of off-copyright newspaper page scans, localizing and OCRing individual content regions like headlines and articles. The headlines, bylines, and article texts that form full articles span multiple bounding boxes - often arranged with complex layouts - and we associate them using a model that combines layout information and language understanding. Then, we use neural methods to accurately predict which articles come from the same underlying source, in the presence of noise and abridgement.
We remove all headline pairs that are below a Levenshtein edit distance, divided by the min length in the pair, of 0.1 from each other, with the aim of removing pairs that are exact duplicates up to OCR noise.
#### Who are the source language producers?
The text data was originally produced by journalists of local U.S. newspapers.
### Annotations
The dataset does not contain any additional annotations.
### Personal and Sensitive Information
The dataset may contain information about individuals, to the extent that this is covered in the headlines of news stories. However we make no additional information about individuals publicly available.
### Data Description
The dataset contains 396,001,930 positive semantic similarity pairs, from 1920 to 1989.
It contains headlines from all 50 states.
## Considerations for Using the Data
### Social Impact of Dataset
The purpose of this dataset is to widen the range of language and topics for training semantic similarity models.
This will facilitate the study of semantic change across space and time.
Specific biases in the dataset are considered in the next section.
### Discussion of Biases
The headlines in the dataset may reflect attitudes and values from the period in which they were written, 1920-1989. This may include instances of racism, sexism and homophobia.
We also note that given that all the newspapers considered are from the U.S., the data is likely to present a Western perspective on the news stories of the day.
### Other Known Limitations
As the dataset is sourced from digitalised text, it contains some OCR errors.
## Additional information
### Licensing Information
HEADLINES is released under the Creative Commons CC-BY 2.0 license.
### Dataset curators
This dataset was created by Emily Silcock and Melissa Dell. For more information, see [Dell Research Harvard](https://dell-research-harvard.github.io/).
### Citation information
Citation coming soon.
|
false | # Dataset Card for "test"
[More Information needed](https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards) |
false | |
false | # Dataset Card for ChEMBL Drug Instruction Tuning
## Dataset Description
- **Homepage:**
- **Repository:**
- **Paper:**
- **Leaderboard:**
- **Point of Contact:**
### Dataset Summary
This dataset card aims to be a base template for new datasets. It has been generated using [this raw template](https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/templates/datasetcard_template.md?plain=1).
### Supported Tasks and Leaderboards
[More Information Needed]
### Languages
[More Information Needed]
## Dataset Structure
### Data Instances
[More Information Needed]
### Data Fields
[More Information Needed]
### Data Splits
[More Information Needed]
## Dataset Creation
### Curation Rationale
[More Information Needed]
### Source Data
#### Initial Data Collection and Normalization
[More Information Needed]
#### Who are the source language producers?
[More Information Needed]
### Annotations
#### Annotation process
[More Information Needed]
#### Who are the annotators?
[More Information Needed]
### Personal and Sensitive Information
[More Information Needed]
## Considerations for Using the Data
### Social Impact of Dataset
[More Information Needed]
### Discussion of Biases
[More Information Needed]
### Other Known Limitations
[More Information Needed]
## Additional Information
### Dataset Curators
[More Information Needed]
### Licensing Information
[More Information Needed]
### Citation Information
[More Information Needed]
### Contributions
[More Information Needed] |
false |
## Dataset Info
I do not own this dataset. This dataset was imported from a mobile app from [**Ministry of Public Works and Transport**](https://play.google.com/store/apps/details?id=com.chanthol.drivingrules)
It's for research purposes only.
The dataset was manually reviewed, but there might still be errors.
## Metrics
Total Duration: 6957.366113 seconds (1.932 hours) |
true | # AutoNLP Dataset for project: song-lyrics
## Table of content
- [Dataset Description](#dataset-description)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
## Dataset Descritpion
This dataset has been automatically processed by AutoNLP for project song-lyrics.
### Languages
The BCP-47 code for the dataset's language is en.
## Dataset Structure
### Data Instances
A sample from this dataset looks as follows:
```json
[
{
"target": 2,
"text": "[Intro: Method Man w/ sample] + (Sunny valentine). We got butter (8X). (The gun'll go the gun'll go.... The gun'll go...). [Raekwon]. Aiyo one thing for sure keep you of all. Keep a nice crib fly away keep to the point. Keep niggaz outta ya face who snakes. Keep bitches in they place keep the mac in a special place. Keep moving for papes keep cool keep doing what you doing. Keep it fly keep me in the crates. Cuz I will erase shit on the real note you'se a waste. It's right here for you I will lace you. Rip you and brace you put a nice W up on ya face. Word to mother you could get chased. It's nothing to taste blood on a thug if he gotta go. All I know is we be giving grace. This is a place from where we make tapes. We make 'em everywhere still in all we be making base. Y'all be making paste these little niggaz they be making shapes. Our shit is art yours is traced. [Chorus: Sunny Valentine]. This is the way that we rolling in the streets. You know when we roll we be packing that heat. The gun'll go the gun'll go the gun'll go the gun'll go. The gun'll go the gun'll go the gun'll go the gun'll go. The gun'll go the gun'll go.... [Method Man]. This is Poverty Island man these animals don't run. Slums where the ambulance don't come. Who got the best base? Fiends waiting to smoke some. Approach something ask him where he getting that coke from. My dudes hug blocks like samurai shogun. Cuz no V and no ones equalling no fun. Who want a treat they know huh? Body to go numb. My woman need funds plus her hair and her toes done. It is what it is though you fuck with the kid flow. That make it hard to get dough the harder to get gold. Harder the piff blow harder when it snow. The pinky and the wrist glow this here what we live for. Get gwop then get low but first thought. We gotta get the work off the gift and the curse boss. Yeah see I'm the shit yo the dirt in the fit no. Hustling from the get-go the motto is get more. [Chorus]. [Masta Killa]. We was quiet flashy brothers strapped all along. With the dirty .38 long twelve hour shift gate. Took case state to state you think he won't hold his weight?. Put ya money on the plate and watch it get scrapped. We get ape up in that club off that juice and Henn. And it's a no win situation fucking with them. You mean like Ewing at the front at the rim finger roll a Dutch. Million dollar stages touched techs gauges bust. Trust no one the lone shogun rugged Timb boot stomper. Damaging lyrical mass destruction launcher. Nothing can calm the quakeage when I break kid. Peace to my brothers up north doing state bids. [Chorus]. [Chorus 2: Sunny Valentine]. Whoa... this is the way we be rolling in the club. You know when we roll we be packing .32 snubs. The gun'll go the gun'll go the gun'll go the gun'll go. The gun'll go the gun'll go the gun'll go the gun'll go. The gun'll go the gun'll go the gun'll go the gun'll go. [Outro: sample to fade]. We got butter..."
},
{
"target": 4,
"text": "[Sean Paul:]. Aye. It's Sean Paul 'long side. The mandem called Jay Sean. Fi di gal dem. Tellin' 'em again what we tell 'em. [Jay Sean:]. Pass me a drink to the left yeah. Said her name was Delilah. And I'm like \"you should come my way\". I already surrender. Damn girl that body's fire. You gon' remember my name. (She should give it up definite). You need it. I need it. We can jump in the deep end. I wanna get lost in your love. I just wanna be close to you. (Just wanna I just wanna). And do all the things you want me to. I just wanna be close to you. (I just wanna I just wanna). And show you the way I feel. You make my love go. You make my love go. You make my love go. In the morning we gon' do it again wake up. I'mma do it like we just broke up and made up. Get up on top of me and work up a sweat work up a sweat. See we can do it any type of way that you want. I'm thinking maybe you're the right kind of wrong. I'm saying baby you won't ever forget my love. You need it. I need it. We can jump in the deep end. I wanna get lost in your love. I just wanna be close to you. (Just wanna I just wanna). And do all the things you want me to. I just wanna be close to you. (I just wanna I just wanna). And show you the way I feel. You make my love go. You make my love go. You make my love go. [Sean Paul:]. Girl mi wan' figure hundred hundred and fifty. Love how you move you know that I'm with it. Perfect size I know that you fit it. Just let me hit it you know mi not quit it. Pon di Dl like Cassie and Diddy. Mi na wound a mi watch we like Sin City. Full time mi run da ting mi tall legend. If you don't come gimme dat would I be offended my girl. Come here down wan' see something me want in life and then waste time. A you a mi pree every day baby full time when ya de pon on mi mind. So mi wine if you give it to me baby girl so we can play. Stick to the ting now I am your king my girl this is what we say. [Jay Sean:]. I just wanna be close to you. (Just wanna I just wanna). And do all the things you want me to. I just wanna be close to you. (I just wanna I just wanna). And show you the way I feel. You make my love go. You make my love go. You make my love go"
}
]
```
### Dataset Fields
The dataset has the following fields (also called "features"):
```json
{
"target": "ClassLabel(num_classes=6, names=['Dance', 'Heavy Metal', 'Hip Hop', 'Indie', 'Pop', 'Rock'], names_file=None, id=None)",
"text": "Value(dtype='string', id=None)"
}
```
### Dataset Splits
This dataset is split into a train and validation split. The split sizes are as follow:
| Split name | Num samples |
| ------------ | ------------------- |
| train | 48493 |
| valid | 5389 |
|
false |
# Dataset Card for MedWiki
## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Repository:** [Github](https://github.com/HazyResearch/medical-ned-integration)
- **Paper:** [Cross-Domain Data Integration for Named Entity Disambiguation in Biomedical Text](https://arxiv.org/abs/2110.08228)
- **Point of Contact:** [Maya Varma](mailto:mvarma2@stanford.edu)
### Dataset Summary
MedWiki is a large sentence dataset collected from a medically-relevant subset of Wikipedia and annotated with biomedical entities in the Unified Medical Language System (UMLS) knowledge base. For each entity, we include a rich set of types sourced from both UMLS and WikiData. Consisting of over 13 million sentences and 17 million entity annotations, MedWiki can be utilized as a pretraining resource for language models and can improve performance of medical named entity recognition and disambiguation systems, especially on rare entities.
Here, we include two configurations of MedWiki (further details in [Dataset Creation](#dataset-creation)):
- `MedWiki-Full` is a large sentence dataset with UMLS medical entity annotations generated through the following two steps: (1) a weak labeling proecedure to annotate WikiData entities in sentences and (2) a data integration approach that maps WikiData entities to their counterparts in UMLS.
- `MedWiki-HQ` is a subset of MedWiki-Full with higher quality labels designed to limit noise that arises from the annotation procedure listed above.
### Languages
The text in the dataset is in English and was obtained from English Wikipedia.
## Dataset Structure
### Data Instances
A typical data point includes a sentence collected from Wikipedia annotated with UMLS medical entities and associated titles and types.
An example from the MedWiki test set looks as follows:
```
{'sent_idx_unq': 57000409,
'sentence': "The hair , teeth , and skeletal side effects of TDO are lifelong , and treatment is used to manage those effects .",
'mentions': ['tdo'],
'entities': ['C2931236'],
'entity_titles': ['Tricho-dento-osseous syndrome 1'],
'types': [['Disease or Syndrome', 'disease', 'rare disease', 'developmental defect during embryogenesis', 'malformation syndrome with odontal and/or periodontal component', 'primary bone dysplasia with increased bone density', 'syndromic hair shaft abnormality']],
'spans': [[10, 11]]}
```
### Data Fields
- `sent_idx_unq`: a unique integer identifier for the data instance
- `sentence`: a string sentence collected from English Wikipedia. Punctuation is separated from words, and the sentence can be tokenized into word-pieces with the .split() method.
- `mentions`: list of medical mentions in the sentence.
- `entities`: list of UMLS medical entity identifiers corresponding to mentions. There is exactly one entity for each mention, and the length of the `entities` list is equal to the length of the `mentions` list.
- `entity_titles`: List of English titles collected from UMLS that describe each entity. The length of the `entity_titles` list is equal to the length of the `entities` list.
- `types`: List of category types associated with each entity, including types collected from UMLS and WikiData.
- `spans`: List of integer pairs representing the word span of each mention in the sentence.
### Data Splits
MedWiki includes two configurations: MedWiki-Full and MedWiki-HQ (described further in [Dataset Creation](#dataset-creation)). For each configuration, data is split into training, development, and test sets. The split sizes are as follow:
| | Train | Dev | Test |
| ----- | ------ | ----- | ---- |
| MedWiki-Full Sentences |11,784,235 | 649,132 | 648,608 |
| MedWiki-Full Mentions |15,981,347 | 876,586 | 877,090 |
| MedWiki-Full Unique Entities | 230,871 | 55,002 | 54,772 |
| MedWiki-HQ Sentences | 2,962,089 | 165,941 | 164,193 |
| MedWiki-HQ Mentions | 3,366,108 | 188,957 | 186,622 |
| MedWiki-HQ Unique Entities | 118,572 | 19,725 | 19,437 |
## Dataset Creation
### Curation Rationale
Existing medical text datasets are generally limited in scope, often obtaining low coverage over the entities and structural resources in the UMLS medical knowledge base. When language models are trained across such datasets, the lack of adequate examples may prevent models from learning the complex reasoning patterns that are necessary for performing effective entity linking or disambiguation, especially for rare entities as shown in prior work by [Orr et al.](http://cidrdb.org/cidr2021/papers/cidr2021_paper13.pdf). Wikipedia, which is often utilized as a rich knowledge source in general text settings, contains references to medical terms and can help address this issue. Here, we curate the MedWiki dataset, which is a large-scale, weakly-labeled dataset that consists of sentences from Wikipedia annotated with medical entities in the UMLS knowledge base. MedWiki can serve as a pretraining dataset for language models and holds potential for improving performance on medical named entity recognition tasks, especially on rare entities.
### Source Data
#### Initial Data Collection and Normalization
MedWiki consists of sentences obtained from the November 2019 dump of English Wikipedia. We split pages into an 80/10/10 train/dev/test split and then segment each page at the sentence-level. This ensures that all sentences associated with a single Wikipedia page are placed in the same split.
#### Who are the source language producers?
The source language producers are editors on English Wikipedia.
### Annotations
#### Annotation process
We create two configurations of our dataset: MedWiki-Full and MedWiki-HQ. We label MedWiki-Full by first annotating all English Wikipedia articles with textual mentions and corresponding WikiData entities; we do so by obtaining gold entity labels from internal page links as well as generating weak labels based on pronouns and alternative entity names (see [Orr et al. 2020](http://cidrdb.org/cidr2021/papers/cidr2021_paper13.pdf) for additional information). Then, we use the off-the-shelf entity linker [Bootleg](https://github.com/HazyResearch/bootleg) to map entities in WikiData to their counterparts in the 2017AA release of the Unified Medical Language System (UMLS), a standard knowledge base for biomedical entities (additional implementation details in forthcoming publication). Any sentence containing at least one UMLS entity is included in MedWiki-Full. We also include types associated with each entity, which are collected from both WikiData and UMLS using the generated UMLS-Wikidata mapping. It is important to note that types obtained from WikiData are filtered according to methods described in [Orr et al. 2020](http://cidrdb.org/cidr2021/papers/cidr2021_paper13.pdf).
Since our labeling procedure introduces some noise into annotations, we also release the MedWiki-HQ dataset configuration with higher-quality labels. To generate MedWiki-HQ, we filtered the UMLS-Wikidata mappings to only include pairs of UMLS medical entities and WikiData items that share a high textual overlap between titles. MedWiki-HQ is a subset of MedWiki-Full.
To evaluate the quality of our UMLS-Wikidata mappings, we find that WikiData includes a small set of "true" labeled mappings between UMLS entities and WikiData items. (Note that we only include WikiData items associated with linked Wikipedia pages.) This set comprises approximately 9.3k UMLS entities in the original UMLS-Wikidata mapping (used for MedWiki-Full) and 5.6k entities in the filtered UMLS-Wikidata mapping (used for MedWiki-HQ). Using these labeled sets, we find that our mapping accuracy is 80.2% for the original UMLS-Wikidata mapping and 94.5% for the filtered UMLS-Wikidata mapping. We also evaluate integration performance on this segment as the proportion of mapped WikiData entities that share a WikiData type with the true entity, suggesting the predicted mapping adds relevant structural resources. Integration performance is 85.4% for the original UMLS-Wikidata mapping and 95.9% for the filtered UMLS-Wikidata mapping. The remainder of items in UMLS have no “true” mappings to WikiData.
#### Who are the annotators?
The dataset was labeled using weak-labeling techniques as described above.
### Personal and Sensitive Information
No personal or sensitive information is included in MedWiki.
## Considerations for Using the Data
### Social Impact of Dataset
The purpose of this dataset is to enable the creation of better named entity recognition systems for biomedical text. MedWiki encompasses a large set of entities in the UMLS knowledge base and includes a rich set of types associated with each entity, which can enable the creation of models that achieve high performance on named entity recognition tasks, especially on rare or unpopular entities. Such systems hold potential for improving automated parsing and information retrieval from large quantities of biomedical text.
### Discussion of Biases
The data included in MedWiki comes from English Wikipedia. Generally, Wikipedia articles are neutral in point of view and aim to avoid bias. However, some [prior work](https://www.hbs.edu/ris/Publication%20Files/15-023_e044cf50-f621-4759-a827-e9a3bf8920c0.pdf) has shown that ideological biases may exist within some Wikipedia articles, especially those that are focused on political issues or those that are written by fewer authors. We anticipate that such biases are rare for medical articles, which are typically comprised of scientific facts. However, it is important to note that bias encoded in Wikipedia is likely to be reflected by MedWiki.
### Other Known Limitations
Since MedWiki was annotated using weak labeling techniques, there is likely some noise in entity annotations. (Note that to address this, we include the MedWiki-HQ configuration, which is a subset of MedWiki-Full with higher quality labels. Additional details in [Dataset Creation](#dataset-creation)).
## Additional Information
### Dataset Curators
MedWiki was curated by Maya Varma, Laurel Orr, Sen Wu, Megan Leszczynski, Xiao Ling, and Chris Ré.
### Licensing Information
Dataset licensed under CC BY 4.0.
### Citation Information
```
@inproceedings{varma-etal-2021-cross-domain,
title = "Cross-Domain Data Integration for Named Entity Disambiguation in Biomedical Text",
author = "Varma, Maya and
Orr, Laurel and
Wu, Sen and
Leszczynski, Megan and
Ling, Xiao and
R{\'e}, Christopher",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2021",
month = nov,
year = "2021",
address = "Punta Cana, Dominican Republic",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.findings-emnlp.388",
pages = "4566--4575",
}
```
### Contributions
Thanks to [@maya124](https://github.com/maya124) for adding this dataset.
|
false |
# Dataset Card for CrowdSpeech
## Dataset Description
- **Repository:** [GitHub](https://github.com/Toloka/CrowdSpeech)
- **Paper:** [Paper](https://openreview.net/forum?id=3_hgF1NAXU7)
- **Point of Contact:** research@toloka.ai
### Dataset Summary
CrowdSpeech is the first publicly available large-scale dataset of crowdsourced audio transcriptions.
The dataset was constructed by annotation [LibriSpeech](https://www.openslr.org/12) on [Toloka crowdsourcing platform](https://toloka.ai).
CrowdSpeech consists of 22K instances having around 155K annotations obtained from crowd workers.
### Supported Tasks and Leaderboards
Aggregation of crowd transcriptions.
### Languages
English
## Dataset Structure
### Data Instances
A data instance contains a url to the audio recording, a list of transcriptions along with the corresponding performers identifiers and ground truth.
For each data instance, seven crowdsourced transcriptions are provided.
```
{'task': 'https://tlk.s3.yandex.net/annotation_tasks/librispeech/train-clean/0.mp3',
'transcriptions': "had laid before her a pair of alternatives now of course you're completely your own mistress and are as free as the bird on the bough i don't mean you were not so before but you're at present on a different footing | had laid before her a pair of alternatives now of course you are completely your own mistress and are as free as the bird on the bowl i don't mean you were not so before but you were present on a different footing | had laid before her a pair of alternatives now of course you're completely your own mistress and are as free as the bird on the bow i don't mean you are not so before but you're at present on a different footing | had laid before her a pair of alternatives now of course you're completely your own mistress and are as free as the bird on the bow i don't mean you are not so before but you're at present on a different footing | laid before her a pair of alternativesnow of course you're completely your own mistress and are as free as the bird on the bow i don't mean you're not so before but you're at present on a different footing | had laid before her a peril alternatives now of course your completely your own mistress and as free as a bird as the back bowl i don't mean you were not so before but you are present on a different footing | a lady before her a pair of alternatives now of course you're completely your own mistress and rs free as the bird on the ball i don't need you or not so before but you're at present on a different footing",
'performers': '1154 | 3449 | 3097 | 461 | 3519 | 920 | 3660',
'gt': "had laid before her a pair of alternatives now of course you're completely your own mistress and are as free as the bird on the bough i don't mean you were not so before but you're at present on a different footing"}
```
### Data Fields
* task: a string containing a url of the audio recording
* transcriptions: a list of the crowdsourced transcriptions separated by '|'
* performers: the corresponding performers' identifiers.
* gt: ground truth transcription
### Data Splits
There are five splits in the data: train, test, test.other, dev.clean and dev.other.
Splits train, test and dev.clean correspond to *clean* part of LibriSpeech that contains audio recordings of higher quality with accents
of the speaker being closer to the US English. Splits dev.other and test.other correspond to *other* part of LibriSpeech with
the recordings more challenging for recognition. The audio recordings are gender-balanced.
## Dataset Creation
### Source Data
[LibriSpeech](https://www.openslr.org/12) is a corpus of approximately 1000 hours of 16kHz read English speech.
### Annotations
Annotation was done on [Toloka crowdsourcing platform](https://toloka.ai) with overlap of 7 (that is, each task was performed by 7 annotators).
Only annotators who self-reported the knowledge of English had access to the annotation task.
Additionally, annotators had to pass *Entrance Exam*. For this, we ask all incoming eligible workers to annotate ten audio
recordings. We then compute our target metric — Word Error Rate (WER) — on these recordings and accept to the main task all workers
who achieve WER of 40% or less (the smaller the value of the metric, the higher the quality of annotation).
The Toloka crowdsourcing platform associates workers with unique identifiers and returns these identifiers to the requester.
To further protect the data, we additionally encode each identifier with an integer that is eventually reported in our released datasets.
See more details in the [paper](https://arxiv.org/pdf/2107.01091.pdf).
### Citation Information
```
@inproceedings{CrowdSpeech,
author = {Pavlichenko, Nikita and Stelmakh, Ivan and Ustalov, Dmitry},
title = {{CrowdSpeech and Vox~DIY: Benchmark Dataset for Crowdsourced Audio Transcription}},
year = {2021},
booktitle = {Proceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks},
eprint = {2107.01091},
eprinttype = {arxiv},
eprintclass = {cs.SD},
url = {https://openreview.net/forum?id=3_hgF1NAXU7},
language = {english},
pubstate = {forthcoming},
}
``` |
true |
# Dataset Card for hate-multi
## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Dataset Creation](#dataset-creation)
- [Source Data](#source-data)
## Dataset Description
### Dataset Summary
This dataset contains a collection of text labeled as hate speech (class 1) or not (class 0).
## Dataset Creation
The dataset was creating by aggregating multiple publicly available datasets.
### Source Data
The following datasets were used:
* https://huggingface.co/datasets/hate_speech18 - Filtered to remove examples labeled as 'idk/skip', 'relation'
* https://huggingface.co/datasets/hate_speech_offensive - Tweet text cleaned by lower casing, removing mentions and urls. Dropped instanced labeled as 'offensive language'
* https://huggingface.co/datasets/ucberkeley-dlab/measuring-hate-speech - Tweet text cleaned by lower casing, removing mentions and urls. Dropped instanced with hatespeech == 1
|
true |
# Dataset Card for news-12factor
## Table of Contents
- [Dataset Description](#dataset-description)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Source Data](#source-data)
- [Annotations](#annotations)
## Dataset Description
80+ news articles with url, title, body text, scored on 12 quality factors and assigned a single rank.
## Languages
The text in the dataset is in English
## Dataset Structure
[Needs More Information]
## Source Data
URL data was scraped using [news-please](https://github.com/fhamborg/news-please)
## Annotations
Articles were manually annotated by Alex on a 12-factor score card.
|
false |
Paper: [Understanding Iterative Revision from Human-Written Text](https://arxiv.org/abs/2203.03802)
Authors: Wanyu Du, Vipul Raheja, Dhruv Kumar, Zae Myung Kim, Melissa Lopez, Dongyeop Kang
Github repo: https://github.com/vipulraheja/IteraTeR
|
false |
Paper: [Understanding Iterative Revision from Human-Written Text](https://arxiv.org/abs/2203.03802)
Authors: Wanyu Du, Vipul Raheja, Dhruv Kumar, Zae Myung Kim, Melissa Lopez, Dongyeop Kang
Github repo: https://github.com/vipulraheja/IteraTeR
|
false | # Dataset Card for Cartoon Set
## Table of Contents
- [Dataset Card for Cartoon Set](#dataset-card-for-cartoon-set)
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Usage](#usage)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Homepage:** https://google.github.io/cartoonset/
- **Repository:** https://github.com/google/cartoonset/
- **Paper:** XGAN: Unsupervised Image-to-Image Translation for Many-to-Many Mappings
- **Leaderboard:**
- **Point of Contact:**
### Dataset Summary
[Cartoon Set](https://google.github.io/cartoonset/) is a collection of random, 2D cartoon avatar images. The cartoons vary in 10 artwork categories, 4 color categories, and 4 proportion categories, with a total of ~10^13 possible combinations. We provide sets of 10k and 100k randomly chosen cartoons and labeled attributes.
#### Usage
`cartoonset` provides the images as PNG byte strings, this gives you a bit more flexibility into how to load the data. Here we show 2 ways:
**Using PIL:**
```python
import datasets
from io import BytesIO
from PIL import Image
ds = datasets.load_dataset("cgarciae/cartoonset", "10k") # or "100k"
def process_fn(sample):
img = Image.open(BytesIO(sample["img_bytes"]))
...
return {"img": img}
ds = ds.map(process_fn, remove_columns=["img_bytes"])
```
**Using TensorFlow:**
```python
import datasets
import tensorflow as tf
hfds = datasets.load_dataset("cgarciae/cartoonset", "10k") # or "100k"
ds = tf.data.Dataset.from_generator(
lambda: hfds,
output_signature={
"img_bytes": tf.TensorSpec(shape=(), dtype=tf.string),
},
)
def process_fn(sample):
img = tf.image.decode_png(sample["img_bytes"], channels=3)
...
return {"img": img}
ds = ds.map(process_fn)
```
**Additional features:**
You can also access the features that generated each sample e.g:
```python
ds = datasets.load_dataset("cgarciae/cartoonset", "10k+features") # or "100k+features"
```
Apart from `img_bytes` these configurations add a total of 18 * 2 additional `int` features, these come in `{feature}`, `{feature}_num_categories` pairs where `num_categories` indicates the number of categories for that feature. See [Data Fields](#data-fields) for the complete list of features.
## Dataset Structure
### Data Instances
A sample from the training set is provided below:
```python
{
'img_bytes': b'0x...',
}
```
If `+features` is added to the dataset name, the following additional fields are provided:
```python
{
'img_bytes': b'0x...',
'eye_angle': 0,
'eye_angle_num_categories': 3,
'eye_lashes': 0,
'eye_lashes_num_categories': 2,
'eye_lid': 0,
'eye_lid_num_categories': 2,
'chin_length': 2,
'chin_length_num_categories': 3,
...
}
```
### Data Fields
- `img_bytes`: A byte string containing the raw data of a 500x500 PNG image.
If `+features` is appended to the dataset name, the following additional `int32` fields are provided:
- `eye_angle`
- `eye_angle_num_categories`
- `eye_lashes`
- `eye_lashes_num_categories`
- `eye_lid`
- `eye_lid_num_categories`
- `chin_length`
- `chin_length_num_categories`
- `eyebrow_weight`
- `eyebrow_weight_num_categories`
- `eyebrow_shape`
- `eyebrow_shape_num_categories`
- `eyebrow_thickness`
- `eyebrow_thickness_num_categories`
- `face_shape`
- `face_shape_num_categories`
- `facial_hair`
- `facial_hair_num_categories`
- `facial_hair_num_categories`
- `facial_hair_num_categories`
- `hair`
- `hair_num_categories`
- `hair_num_categories`
- `hair_num_categories`
- `eye_color`
- `eye_color_num_categories`
- `face_color`
- `face_color_num_categories`
- `hair_color`
- `hair_color_num_categories`
- `glasses`
- `glasses_num_categories`
- `glasses_color`
- `glasses_color_num_categories`
- `eyes_slant`
- `eye_slant_num_categories`
- `eyebrow_width`
- `eyebrow_width_num_categories`
- `eye_eyebrow_distance`
- `eye_eyebrow_distance_num_categories`
### Data Splits
Train
## Dataset Creation
### Licensing Information
This data is licensed by Google LLC under a Creative Commons Attribution 4.0 International License.
### Citation Information
```
@article{DBLP:journals/corr/abs-1711-05139,
author = {Amelie Royer and
Konstantinos Bousmalis and
Stephan Gouws and
Fred Bertsch and
Inbar Mosseri and
Forrester Cole and
Kevin Murphy},
title = {{XGAN:} Unsupervised Image-to-Image Translation for many-to-many Mappings},
journal = {CoRR},
volume = {abs/1711.05139},
year = {2017},
url = {http://arxiv.org/abs/1711.05139},
eprinttype = {arXiv},
eprint = {1711.05139},
timestamp = {Mon, 13 Aug 2018 16:47:38 +0200},
biburl = {https://dblp.org/rec/journals/corr/abs-1711-05139.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
### Contributions
|
true | # AutoNLP Dataset for project: tweet-sentiment
## Table of content
- [Dataset Description](#dataset-description)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
## Dataset Descritpion
This dataset has been automatically processed by AutoNLP for project tweet-sentiment.
### Languages
The BCP-47 code for the dataset's language is en.
## Dataset Structure
### Data Instances
A sample from this dataset looks as follows:
```json
[
{
"text": "I am going to see how long I can do this for.",
"target": 8
},
{
"text": "@anitabora yeah, right. What if our politicians start using uploading their pics, lots of inside sto[...]",
"target": 8
}
]
```
### Dataset Fields
The dataset has the following fields (also called "features"):
```json
{
"text": "Value(dtype='string', id=None)",
"target": "ClassLabel(num_classes=13, names=['anger', 'boredom', 'empty', 'enthusiasm', 'fun', 'happiness', 'hate', 'love', 'neutral', 'relief', 'sadness', 'surprise', 'worry'], id=None)"
}
```
### Dataset Splits
This dataset is split into a train and validation split. The split sizes are as follow:
| Split name | Num samples |
| ------------ | ------------------- |
| train | 31995 |
| valid | 8005 |
|
false |
# Dataset Card for [Dataset Name]
## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Homepage:**
- **Repository:**
- **Paper:**
- **Leaderboard:**
- **Point of Contact:**
### Dataset Summary
[More Information Needed]
### Supported Tasks and Leaderboards
[More Information Needed]
### Languages
[More Information Needed]
## Dataset Structure
### Data Instances
[More Information Needed]
### Data Fields
[More Information Needed]
### Data Splits
[More Information Needed]
## Dataset Creation
### Curation Rationale
[More Information Needed]
### Source Data
#### Initial Data Collection and Normalization
[More Information Needed]
#### Who are the source language producers?
[More Information Needed]
### Annotations
#### Annotation process
[More Information Needed]
#### Who are the annotators?
[More Information Needed]
### Personal and Sensitive Information
[More Information Needed]
## Considerations for Using the Data
### Social Impact of Dataset
[More Information Needed]
### Discussion of Biases
[More Information Needed]
### Other Known Limitations
[More Information Needed]
## Additional Information
### Dataset Curators
[More Information Needed]
### Licensing Information
[More Information Needed]
### Citation Information
[More Information Needed]
### Contributions
Thanks to [@github-username](https://github.com/<github-username>) for adding this dataset.
|
false |
# Dataset Card for [Dataset Name]
## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Homepage:**
- **Repository:**
- **Paper:**
- **Leaderboard:**
- **Point of Contact:**
### Dataset Summary
[More Information Needed]
### Supported Tasks and Leaderboards
[More Information Needed]
### Languages
[More Information Needed]
## Dataset Structure
### Data Instances
[More Information Needed]
### Data Fields
[More Information Needed]
### Data Splits
[More Information Needed]
## Dataset Creation
### Curation Rationale
[More Information Needed]
### Source Data
#### Initial Data Collection and Normalization
[More Information Needed]
#### Who are the source language producers?
[More Information Needed]
### Annotations
#### Annotation process
[More Information Needed]
#### Who are the annotators?
[More Information Needed]
### Personal and Sensitive Information
[More Information Needed]
## Considerations for Using the Data
### Social Impact of Dataset
[More Information Needed]
### Discussion of Biases
[More Information Needed]
### Other Known Limitations
[More Information Needed]
## Additional Information
### Dataset Curators
[More Information Needed]
### Licensing Information
[More Information Needed]
### Citation Information
[More Information Needed]
### Contributions
Thanks to [@github-username](https://github.com/<github-username>) for adding this dataset.
|
false |
This is a suite of psycholinguistic datasets by Allyson Ettinger. See her [official Github repository](https://github.com/aetting/lm-diagnostics) for specific details. |
false |
# Spanish to Quechua
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Structure](#dataset-structure)
- [Dataset Creation](#dataset-creation)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [team members](#team-members)
## Dataset Description
This dataset is a recopilation of webs and others datasets that shows in [dataset creation section](#dataset-creation). This contains translations from spanish (es) to Qechua of Ayacucho (qu).
## Dataset Structure
### Data Fields
- es: The sentence in Spanish.
- qu: The sentence in Quechua of Ayacucho.
### Data Splits
- train: To train the model (102 747 sentences).
- Validation: To validate the model during training (12 844 sentences).
- test: To evaluate the model when the training is finished (12 843 sentences).
## Dataset Creation
### Source Data
This dataset has generated from:
- "Mundo Quechua" by "Ivan Acuña" - [available here](https://mundoquechua.blogspot.com/2006/07/frases-comunes-en-quechua.html)
- "Kuyakuykim (Te quiero): Apps con las que podrías aprender quechua" by "El comercio" - [available here](https://elcomercio.pe/tecnologia/actualidad/traductor-frases-romanticas-quechua-noticia-467022-noticia/)
- "Piropos y frases de amor en quechua" by "Soy Quechua" - [available here](https://www.soyquechua.org/2019/12/palabras-en-quechua-de-amor.html)
- "Corazón en quechua" by "Soy Quechua" - [available here](https://www.soyquechua.org/2020/05/corazon-en-quechua.html)
- "Oraciones en Español traducidas a Quechua" by "Tatoeba" - [available here](https://tatoeba.org/es/sentences/search?from=spa&query=&to=que)
- "AmericasNLP 2021 Shared Task on Open Machine Translation" by "americasnlp2021" - [available here](https://github.com/AmericasNLP/americasnlp2021/tree/main/data/quechua-spanish/parallel_data/es-quy)
### Data cleaning
- The dataset was manually cleaned during compilation, as some words of one language were related to several words of the other language.
## Considerations for Using the Data
This is a first version of the dataset, we expected improve it over time and especially to neutralize the biblical themes.
## Team members
- [Sara Benel](https://huggingface.co/sbenel)
- [Jose Vílchez](https://huggingface.co/JCarlos) |
false |
# TermITH-Eval Benchmark Dataset for Keyphrase Generation
## About
TermITH-Eval is a dataset for benchmarking keyphrase extraction and generation models.
The dataset is composed of 400 abstracts of scientific papers in French collected from the FRANCIS and PASCAL databases of the French [Institute for Scientific and Technical Information (Inist)](https://www.inist.fr/).
Keyphrases were annotated by professional indexers in an uncontrolled setting (that is, not limited to thesaurus entries).
Details about the dataset can be found in the original paper [(Bougouin et al., 2016)][bougouin-2016].
Reference (indexer-assigned) keyphrases are also categorized under the PRMU (<u>P</u>resent-<u>R</u>eordered-<u>M</u>ixed-<u>U</u>nseen) scheme as proposed in [(Boudin and Gallina, 2021)][boudin-2021]. Present reference keyphrases are also ordered by their order of apparition in the concatenation of title and abstract.
Text pre-processing (tokenization) is carried out using `spacy` (`fr_core_news_sm` model) with a special rule to avoid splitting words with hyphens (e.g. graph-based is kept as one token).
Stemming (Snowball stemmer implementation for french provided in `nltk`) is applied before reference keyphrases are matched against the source text.
Details about the process can be found in `prmu.py`.
## Content and statistics
The dataset contains the following test split:
| Split | # documents | #words | # keyphrases | % Present | % Reordered | % Mixed | % Unseen |
| :--------- |------------:|-----------:|-------------:|----------:|------------:|--------:|---------:|
| Test | 399 | 156.9 | 11.81 | 40.60 | 7.32 | 19.28 | 32.80 |
The following data fields are available :
- **id**: unique identifier of the document.
- **title**: title of the document.
- **abstract**: abstract of the document.
- **keyphrases**: list of reference keyphrases.
- **prmu**: list of <u>P</u>resent-<u>R</u>eordered-<u>M</u>ixed-<u>U</u>nseen categories for reference keyphrases.
- **category**: category of the document, i.e. chimie (chemistry), archeologie (archeology), linguistique (linguistics) and scienceInfo (information sciences).
## References
- (Bougouin et al., 2016) Adrien Bougouin, Sabine Barreaux, Laurent Romary, Florian Boudin, and Béatrice Daille. 2016.
[TermITH-Eval: a French Standard-Based Resource for Keyphrase Extraction Evaluation][bougouin-2016].
In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC'16), pages 1924–1927, Portorož, Slovenia. European Language Resources Association (ELRA).Language Processing, pages 543–551, Nagoya, Japan. Asian Federation of Natural Language Processing.
- (Boudin and Gallina, 2021) Florian Boudin and Ygor Gallina. 2021.
[Redefining Absent Keyphrases and their Effect on Retrieval Effectiveness][boudin-2021].
In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 4185–4193, Online. Association for Computational Linguistics.
[bougouin-2016]: https://aclanthology.org/L16-1304/
[boudin-2021]: https://aclanthology.org/2021.naacl-main.330/ |
false |
# Dataset Card for broad_twitter_corpus
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-instances)
- [Data Splits](#data-instances)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
## Dataset Description
- **Homepage:** [https://github.com/GateNLP/broad_twitter_corpus](https://github.com/GateNLP/broad_twitter_corpus)
- **Repository:** [https://github.com/GateNLP/broad_twitter_corpus](https://github.com/GateNLP/broad_twitter_corpus)
- **Paper:** [http://www.aclweb.org/anthology/C16-1111](http://www.aclweb.org/anthology/C16-1111)
- **Leaderboard:** [Named Entity Recognition on Broad Twitter Corpus](https://paperswithcode.com/sota/named-entity-recognition-on-broad-twitter)
- **Point of Contact:** [Leon Derczynski](https://github.com/leondz)
### Dataset Summary
This is the Broad Twitter corpus, a dataset of tweets collected over stratified times, places and social uses. The goal is to represent a broad range of activities, giving a dataset more representative of the language used in this hardest of social media formats to process. Further, the BTC is annotated for named entities.
See the paper, [Broad Twitter Corpus: A Diverse Named Entity Recognition Resource](http://www.aclweb.org/anthology/C16-1111), for details.
### Supported Tasks and Leaderboards
* Named Entity Recognition
* On PWC: [Named Entity Recognition on Broad Twitter Corpus](https://paperswithcode.com/sota/named-entity-recognition-on-broad-twitter)
### Languages
English from UK, US, Australia, Canada, Ireland, New Zealand; `bcp47:en`
## Dataset Structure
### Data Instances
Feature |Count
---|---:
Documents |9 551
Tokens |165 739
Person entities |5 271
Location entities |3 114
Organization entities |3 732
### Data Fields
Each tweet contains an ID, a list of tokens, and a list of NER tags
- `id`: a `string` feature.
- `tokens`: a `list` of `strings`
- `ner_tags`: a `list` of class IDs (`int`s) representing the NER class:
```
0: O
1: B-PER
2: I-PER
3: B-ORG
4: I-ORG
5: B-LOC
6: I-LOC
```
### Data Splits
Section|Region|Collection period|Description|Annotators|Tweet count
---|---|---|---|---|---:
A | UK| 2012.01| General collection |Expert| 1000
B |UK |2012.01-02 |Non-directed tweets |Expert |2000
E |Global| 2014.07| Related to MH17 disaster| Crowd & expert |200
F |Stratified |2009-2014| Twitterati |Crowd & expert |2000
G |Stratified| 2011-2014| Mainstream news| Crowd & expert| 2351
H |Non-UK| 2014 |General collection |Crowd & expert |2000
The most varied parts of the BTC are sections F and H. However, each of the remaining four sections has some specific readily-identifiable bias. So, we propose that one uses half of section H for evaluation and leaves the other half in the training data. Section H should be partitioned in the order of the JSON-format lines. Note that the CoNLL-format data is readily reconstructible from the JSON format, which is the authoritative data format from which others are derived.
**Test**: Section F
**Development**: Section H (the paper says "second half of Section H" but ordinality could be ambiguous, so it all goes in. Bonne chance)
**Training**: everything else
## Dataset Creation
### Curation Rationale
[Needs More Information]
### Source Data
#### Initial Data Collection and Normalization
[Needs More Information]
#### Who are the source language producers?
[Needs More Information]
### Annotations
#### Annotation process
[Needs More Information]
#### Who are the annotators?
[Needs More Information]
### Personal and Sensitive Information
[Needs More Information]
## Considerations for Using the Data
### Social Impact of Dataset
[Needs More Information]
### Discussion of Biases
[Needs More Information]
### Other Known Limitations
[Needs More Information]
## Additional Information
### Dataset Curators
[Needs More Information]
### Licensing Information
Creative Commons Attribution 4.0 International (CC BY 4.0)
### Citation Information
```
@inproceedings{derczynski2016broad,
title={Broad twitter corpus: A diverse named entity recognition resource},
author={Derczynski, Leon and Bontcheva, Kalina and Roberts, Ian},
booktitle={Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers},
pages={1169--1179},
year={2016}
}
```
### Contributions
Author-added dataset [@leondz](https://github.com/leondz)
|
false |
# Dataset Card for "shaj"
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Homepage:**
- **Repository:** [https://figshare.com/articles/dataset/SHAJ_Albanian_hate_speech_abusive_language/19333298/1](https://figshare.com/articles/dataset/SHAJ_Albanian_hate_speech_abusive_language/19333298/1)
- **Paper:** [https://arxiv.org/abs/2107.13592](https://arxiv.org/abs/2107.13592)
- **Point of Contact:** [Leon Derczynski](https://github.com/leondz)
- **Size of downloaded dataset files:** 769.21 KiB
- **Size of the generated dataset:** 1.06 MiB
- **Total amount of disk used:** 1.85 MiB
### Dataset Summary
This is an abusive/offensive language detection dataset for Albanian. The data is formatted
following the OffensEval convention, with three tasks:
* Subtask A: Offensive (OFF) or not (NOT)
* Subtask B: Untargeted (UNT) or targeted insult (TIN)
* Subtask C: Type of target: individual (IND), group (GRP), or other (OTH)
Notes on the above:
* The subtask A field should always be filled.
* The subtask B field should only be filled if there's "offensive" (OFF) in A.
* The subtask C field should only be filled if there's "targeted" (TIN) in B.
The dataset name is a backronym, also standing for "Spoken Hate in the Albanian Jargon"
See the paper [https://arxiv.org/abs/2107.13592](https://arxiv.org/abs/2107.13592) for full details.
### Supported Tasks and Leaderboards
* Task A leaderboard at [paperswithcode.com/sota/hate-speech-detection-on-shaj](https://paperswithcode.com/sota/hate-speech-detection-on-shaj)
### Languages
Albanian (`bcp47:sq-AL`)
## Dataset Structure
### Data Instances
#### shaj
- **Size of downloaded dataset files:** 769.21 KiB
- **Size of the generated dataset:** 1.06 MiB
- **Total amount of disk used:** 1.85 MiB
An example of 'train' looks as follows.
```
{
'id': '0',
'text': 'PLACEHOLDER TEXT',
'subtask_a': 1,
'subtask_b': 0,
'subtask_c': 0
}
```
### Data Fields
- `id`: a `string` feature.
- `text`: a `string`.
- `subtask_a`: whether or not the instance is offensive; `0: OFF, 1: NOT`
- `subtask_b`: whether an offensive instance is a targeted insult; `0: TIN, 1: UNT, 2: not applicable`
- `subtask_c`: what a targeted insult is aimed at; `0: IND, 1: GRP, 2: OTH, 3: not applicable`
### Data Splits
| name |train|
|---------|----:|
|shaj|11874 sentences|
## Dataset Creation
### Curation Rationale
Collecting data for enabling offensive speech detection in Albanian
### Source Data
#### Initial Data Collection and Normalization
The text is scraped from comments on popular Albanian YouTube and Instagram accounts.
An extended discussion is given in the paper in section 3.2.
#### Who are the source language producers?
People who comment on a selection of high-activity Albanian instagram and youtube profiles.
### Annotations
#### Annotation process
The annotation scheme was taken from OffensEval 2019 and applied by two native speaker authors of the paper as well as their friends and family.
#### Who are the annotators?
Albanian native speakers, male and female, aged 20-60.
### Personal and Sensitive Information
The data was public at the time of collection. No PII removal has been performed.
## Considerations for Using the Data
### Social Impact of Dataset
The data definitely contains abusive language.
### Discussion of Biases
### Other Known Limitations
## Additional Information
### Dataset Curators
The dataset is curated by the paper's authors.
### Licensing Information
The authors distribute this data under Creative Commons attribution license, CC-BY 4.0.
### Citation Information
```
@article{nurce2021detecting,
title={Detecting Abusive Albanian},
author={Nurce, Erida and Keci, Jorgel and Derczynski, Leon},
journal={arXiv preprint arXiv:2107.13592},
year={2021}
}
```
### Contributions
Author-added dataset [@leondz](https://github.com/leondz)
|
false |
# Ukrainian Wikipedia Edits
### Dataset summary
A collection of over 5M sentence edits extracted from Ukrainian Wikipedia history revisions.
Edits were filtered by edit distance and sentence length. This makes them usable for grammatical error correction (GEC) or spellchecker models pre-training.
### Supported Tasks and Leaderboards
* Ukrainian grammatical error correction (GEC) - see [UA-GEC](https://github.com/grammarly/ua-gec)
* Ukrainian spelling correction
### Languages
Ukrainian
## Dataset Structure
### Data Fields
* `src` - sentence before edit
* `tgt` - sentence after edit
### Data Splits
* `full/train` contains all the data (5,243,376 samples)
* `tiny/train` contains a 5000 examples sample.
## Dataset Creation
Latest full Ukrainian Wiki dump were used as of 2022-04-30.
It was processed with the [wikiedits](https://github.com/snukky/wikiedits) and custom scripts.
### Source Data
#### Initial Data Collection and Normalization
Wikipedia
#### Who are the source language producers?
Wikipedia writers
### Annotations
#### Annotation process
Annotations inferred by comparing two subsequent page revisions.
#### Who are the annotators?
People who edit Wikipedia pages.
### Personal and Sensitive Information
No
## Considerations for Using the Data
### Social Impact of Dataset
[More Information Needed]
### Discussion of Biases
[More Information Needed]
### Other Known Limitations
The data is noisy. In addition to GEC and spelling edits, it contains a good chunk of factual changes and vandalism.
More task-specific filters could help.
## Additional Information
### Dataset Curators
[Oleksiy Syvokon](https://github.com/asivokon)
### Licensing Information
CC-BY-3.0
### Citation Information
```
@inproceedings{wiked2014,
author = {Roman Grundkiewicz and Marcin Junczys-Dowmunt},
title = {The WikEd Error Corpus: A Corpus of Corrective Wikipedia Edits and its Application to Grammatical Error Correction},
booktitle = {Advances in Natural Language Processing -- Lecture Notes in Computer Science},
editor = {Adam Przepiórkowski and Maciej Ogrodniczuk},
publisher = {Springer},
year = {2014},
volume = {8686},
pages = {478--490},
url = {http://emjotde.github.io/publications/pdf/mjd.poltal2014.draft.pdf}
}
```
### Contributions
[@snukky](https://github.com/snukky) created tools for dataset processing.
[@asivokon](https://github.com/asivokon) generated this dataset.
|
true | # Dataset Card for OSDG-CD
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Homepage:** [OSDG-CD homepage](https://doi.org/10.5281/zenodo.6393942)
### Dataset Summary
The OSDG Community Dataset (OSDG-CD) is a public dataset of thousands of text excerpts, which were validated by approximately 1,000 OSDG Community Platform (OSDG-CP) citizen scientists from over 110 countries, with respect to the Sustainable Development Goals (SDGs).
> **_NOTE:_** There are currently no examples for SDGs 16 and 17. See [this GitHub issue](https://github.com/osdg-ai/osdg-data/issues/3).
### Supported Tasks and Leaderboards
TBD
### Languages
The language of the dataset is English.
## Dataset Structure
### Data Instances
For each instance, there is a string for the text, a string for the SDG, and an integer for the label.
```
{'text': 'Each section states the economic principle, reviews international good practice and discusses the situation in Brazil.',
'label': 5}
```
The average token count for the premises and hypotheses are given below:
| Feature | Mean Token Count |
| ---------- | ---------------- |
| Premise | 14.1 |
| Hypothesis | 8.3 |
### Data Fields
- `doi`: Digital Object Identifier of the original document
- `text_id`: unique text identifier
- `text`: text excerpt from the document
- `sdg`: the SDG the text is validated against
- `label`: an integer from `0` to `17` which corresponds to the `sdg` field
- `labels_negative`: the number of volunteers who rejected the suggested SDG label
- `labels_positive`: the number of volunteers who accepted the suggested SDG label
- `agreement`: agreement score based on the formula
### Data Splits
The OSDG-CD dataset has 1 splits: _train_.
| Dataset Split | Number of Instances in Split |
| ------------- |----------------------------- |
| Train | 32,327 |
## Dataset Creation
### Curation Rationale
The [The OSDG Community Dataset (OSDG-CD)](https://doi.org/10.5281/zenodo.6393942) was developed as a benchmark for ...
with the goal of producing a dataset large enough to train models using neural methodologies.
### Source Data
#### Initial Data Collection and Normalization
TBD
#### Who are the source language producers?
TBD
### Annotations
#### Annotation process
TBD
#### Who are the annotators?
TBD
### Personal and Sensitive Information
The dataset does not contain any personal information about the authors or the crowdworkers.
## Considerations for Using the Data
### Social Impact of Dataset
TBD
## Additional Information
TBD
### Dataset Curators
TBD
### Licensing Information
The OSDG Community Dataset (OSDG-CD) is licensed under a [Creative Commons Attribution 4.0 International License](http://creativecommons.org/licenses/by/4.0/).
### Citation Information
```
@dataset{osdg_2022_6393942,
author = {OSDG and
UNDP IICPSD SDG AI Lab and
PPMI},
title = {OSDG Community Dataset (OSDG-CD)},
month = apr,
year = 2022,
note = {{This CSV file uses UTF-8 character encoding. For
easy access on MS Excel, open the file using Data
→ From Text/CSV. Please split CSV data into
different columns by using a TAB delimiter.}},
publisher = {Zenodo},
version = {2022.04},
doi = {10.5281/zenodo.6393942},
url = {https://doi.org/10.5281/zenodo.6393942}
}
```
### Contributions
TBD
|
false |
# Dataset Card for ASCEND
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-instances)
- [Data Splits](#data-instances)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
## Dataset Description
- **Homepage:** [Needs More Information]
- **Repository:** [Needs More Information]
- **Paper:** https://arxiv.org/abs/2112.06223
- **Leaderboard:** [Needs More Information]
- **Point of Contact:** [Needs More Information]
### Dataset Summary
ASCEND (A Spontaneous Chinese-English Dataset) introduces a high-quality resource of spontaneous multi-turn conversational dialogue Chinese-English code-switching corpus collected in Hong Kong. ASCEND consists of 10.62 hours of spontaneous speech with a total of ~12.3K utterances. The corpus is split into 3 sets: training, validation, and test with a ratio of 8:1:1 while maintaining a balanced gender proportion on each set.
### Supported Tasks and Leaderboards
[Needs More Information]
### Languages
Chinese and English
## Dataset Structure
### Data Instances
[Needs More Information]
### Data Fields
[Needs More Information]
### Data Splits
[Needs More Information]
## Dataset Creation
### Curation Rationale
[Needs More Information]
### Source Data
#### Initial Data Collection and Normalization
[Needs More Information]
#### Who are the source language producers?
[Needs More Information]
### Annotations
#### Annotation process
[Needs More Information]
#### Who are the annotators?
[Needs More Information]
### Personal and Sensitive Information
[Needs More Information]
## Considerations for Using the Data
### Social Impact of Dataset
[Needs More Information]
### Discussion of Biases
[Needs More Information]
### Other Known Limitations
[Needs More Information]
## Additional Information
### Dataset Curators
[Needs More Information]
### Licensing Information
[Needs More Information]
### Citation Information
[Needs More Information] |
false |
# Dataset Card for Team-PIXEL/rendered-wikipedia-english
## Dataset Description
- **Homepage:** [https://github.com/xplip/pixel](https://github.com/xplip/pixel)
- **Repository:** [https://github.com/xplip/pixel](https://github.com/xplip/pixel)
- **Paper:** [Language Modelling with Pixels](https://arxiv.org/abs/2207.06991)
- **Point of Contact:** [Phillip Rust](mailto:p.rust@di.ku.dk)
- **Size of downloaded dataset files:** 125.66 GB
- **Size of the generated dataset:** 125.56 GB
- **Total amount of disk used:** 251.22 GB
### Dataset Summary
This dataset contains the full English Wikipedia from February 1, 2018, rendered into images of 16x8464 resolution.
The original text dataset was built from a [Wikipedia dump](https://dumps.wikimedia.org/). Each example in the original *text* dataset contained the content of one full Wikipedia article with cleaning to strip markdown and unwanted sections (references, etc.). Each *rendered* example contains a subset of one full article. This rendered English Wikipedia was used to train the [PIXEL](https://huggingface.co/Team-PIXEL/pixel-base) model introduced in the paper [Language Modelling with Pixels](https://arxiv.org/abs/2207.06991) by Phillip Rust, Jonas F. Lotz, Emanuele Bugliarello, Elizabeth Salesky, Miryam de Lhoneux, and Desmond Elliott.
The original Wikipedia text dataset was rendered article-by-article into 11.4M examples containing approximately 2B words in total. The dataset is stored as a collection of 338 parquet files.
It was rendered using the script openly available at [https://github.com/xplip/pixel/blob/main/scripts/data/prerendering/prerender_wikipedia.py](https://github.com/xplip/pixel/blob/main/scripts/data/prerendering/prerender_wikipedia.py). The text renderer uses a PyGame backend and a collection of merged Google Noto Sans fonts. The PyGame backend does not support complex text layouts (e.g. ligatures and right-to-left scripts) or emoji, so occurrences of such text in the Wikipedia data have not been rendered accurately.
Each example consists of a "pixel_values" field which stores a 16x8464 (height, width) grayscale image containing the rendered text, and an integer value "num_patches" which stores how many image patches (when splitting the image into 529 non-overlapping patches of resolution 16x16 pixels) in the associated images contain actual text, i.e. are neither blank (fully white) nor are the fully black end-of-sequence patch.
You can load the dataset as follows:
```python
from datasets import load_dataset
# Download the full dataset to disk
load_dataset("Team-PIXEL/rendered-wikipedia-english", split="train")
# Stream the dataset directly from the hub
load_dataset("Team-PIXEL/rendered-wikipedia-english", split="train", streaming=True)
```
## Dataset Structure
### Data Instances
- **Size of downloaded dataset files:** 125.66 GB
- **Size of the generated dataset:** 125.56 GB
- **Total amount of disk used:** 251.22 GB
An example of 'train' looks as follows.
```
{
"pixel_values": <PIL.PngImagePlugin.PngImageFile image mode=L size=8464x16
"num_patches": "469"
}
```
### Data Fields
The data fields are the same among all splits.
- `pixel_values`: an `Image` feature.
- `num_patches`: a `Value(dtype="int64")` feature.
### Data Splits
|train|
|:----|
|11446535|
## Dataset Creation
### Curation Rationale
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Source Data
#### Initial Data Collection and Normalization
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
#### Who are the source language producers?
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Annotations
#### Annotation process
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
#### Who are the annotators?
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Personal and Sensitive Information
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
## Considerations for Using the Data
### Social Impact of Dataset
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Discussion of Biases
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Other Known Limitations
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
## Additional Information
### Dataset Curators
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Licensing Information
Most of Wikipedia's text and many of its images are co-licensed under the Creative Commons Attribution-ShareAlike 3.0 Unported License (CC BY-SA) and the GNU Free Documentation License (GFDL) (unversioned, with no invariant sections, front-cover texts, or back-cover texts).
Some text has been imported only under CC BY-SA and CC BY-SA-compatible license and cannot be reused under GFDL; such text will be identified on the page footer, in the page history, or on the discussion page of the article that utilizes the text.
### Citation Information
```bibtex
@article{rust-etal-2022-pixel,
title={Language Modelling with Pixels},
author={Phillip Rust and Jonas F. Lotz and Emanuele Bugliarello and Elizabeth Salesky and Miryam de Lhoneux and Desmond Elliott},
journal={arXiv preprint},
year={2022},
url={https://arxiv.org/abs/2207.06991}
}
```
### Contact Person
This dataset was added by Phillip Rust.
Github: [@xplip](https://github.com/xplip)
Twitter: [@rust_phillip](https://twitter.com/rust_phillip) |
false |
# Dataset Card for [Dataset Name]
## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Homepage:** http://raingo.github.io/TGIF-Release/
- **Repository:** https://github.com/raingo/TGIF-Release
- **Paper:** https://arxiv.org/abs/1604.02748
- **Point of Contact:** mailto: yli@cs.rochester.edu
### Dataset Summary
The Tumblr GIF (TGIF) dataset contains 100K animated GIFs and 120K sentences describing visual content of the animated GIFs. The animated GIFs have been collected from Tumblr, from randomly selected posts published between May and June of 2015. We provide the URLs of animated GIFs in this release. The sentences are collected via crowdsourcing, with a carefully designed annotation interface that ensures high quality dataset. We provide one sentence per animated GIF for the training and validation splits, and three sentences per GIF for the test split. The dataset shall be used to evaluate animated GIF/video description techniques.
### Languages
The captions in the dataset are in English.
## Dataset Structure
### Data Fields
- `video_path`: `str` "https://31.media.tumblr.com/001a8b092b9752d260ffec73c0bc29cd/tumblr_ndotjhRiX51t8n92fo1_500.gif"
-`video_bytes`: `large_bytes` video file in bytes format
- `en_global_captions`: `list_str` List of english captions describing the entire video
### Data Splits
| |train |validation| test | Overall |
|-------------|------:|---------:|------:|------:|
|# of GIFs|80,000 |10,708 |11,360 |102,068 |
### Annotations
Quoting [TGIF paper](https://arxiv.org/abs/1604.02748): \
"We annotated animated GIFs with natural language descriptions using the crowdsourcing service CrowdFlower.
We carefully designed our annotation task with various
quality control mechanisms to ensure the sentences are both
syntactically and semantically of high quality.
A total of 931 workers participated in our annotation
task. We allowed workers only from Australia, Canada, New Zealand, UK and USA in an effort to collect fluent descriptions from native English speakers. Figure 2 shows the
instructions given to the workers. Each task showed 5 animated GIFs and asked the worker to describe each with one
sentence. To promote language style diversity, each worker
could rate no more than 800 images (0.7% of our corpus).
We paid 0.02 USD per sentence; the entire crowdsourcing
cost less than 4K USD. We provide details of our annotation
task in the supplementary material."
### Personal and Sensitive Information
Nothing specifically mentioned in the paper.
## Considerations for Using the Data
### Social Impact of Dataset
[More Information Needed]
### Discussion of Biases
[More Information Needed]
### Other Known Limitations
[More Information Needed]
## Additional Information
### Licensing Information
This dataset is provided to be used for approved non-commercial research purposes. No personally identifying information is available in this dataset.
### Citation Information
```bibtex
@InProceedings{tgif-cvpr2016,
author = {Li, Yuncheng and Song, Yale and Cao, Liangliang and Tetreault, Joel and Goldberg, Larry and Jaimes, Alejandro and Luo, Jiebo},
title = "{TGIF: A New Dataset and Benchmark on Animated GIF Description}",
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2016}
}
```
### Contributions
Thanks to [@leot13](https://github.com/leot13) for adding this dataset.
|
true | # Dataset Card for sd-nlp
## Table of Contents
- [Dataset Card for [EMBO/sd-nlp-non-tokenized]](#dataset-card-for-dataset-name)
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Initial Data Collection and Normalization](#initial-data-collection-and-normalization)
- [Who are the source language producers?](#who-are-the-source-language-producers)
- [Annotations](#annotations)
- [Annotation process](#annotation-process)
- [Who are the annotators?](#who-are-the-annotators)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Homepage:** https://sourcedata.embo.org
- **Repository:** https://github.com/source-data/soda-roberta
- **Paper:**
- **Leaderboard:**
- **Point of Contact:** thomas.lemberger@embo.org, jorge.abreu@embo.org
### Dataset Summary
This dataset is based on the content of the SourceData (https://sourcedata.embo.org) database, which contains manually annotated figure legends written in English and extracted from scientific papers in the domain of cell and molecular biology (Liechti et al, Nature Methods, 2017, https://doi.org/10.1038/nmeth.4471).
Unlike the dataset [`sd-nlp`](https://huggingface.co/datasets/EMBO/sd-nlp), pre-tokenized with the `roberta-base` tokenizer, this dataset is not previously tokenized, but just splitted into words. Users can therefore use it to fine-tune other models.
Additional details at https://github.com/source-data/soda-roberta
### Supported Tasks and Leaderboards
Tags are provided as [IOB2-style tags](https://en.wikipedia.org/wiki/Inside%E2%80%93outside%E2%80%93beginning_(tagging)).
`PANELIZATION`: figure captions (or figure legends) are usually composed of segments that each refer to one of several 'panels' of the full figure. Panels tend to represent results obtained with a coherent method and depicts data points that can be meaningfully compared to each other. `PANELIZATION` provide the start (B-PANEL_START) of these segments and allow to train for recogntion of the boundary between consecutive panel lengends.
`NER`: biological and chemical entities are labeled. Specifically the following entities are tagged:
- `SMALL_MOLECULE`: small molecules
- `GENEPROD`: gene products (genes and proteins)
- `SUBCELLULAR`: subcellular components
- `CELL`: cell types and cell lines.
- `TISSUE`: tissues and organs
- `ORGANISM`: species
- `DISEASE`: diseases (see limitations)
- `EXP_ASSAY`: experimental assays
`ROLES`: the role of entities with regard to the causal hypotheses tested in the reported results. The tags are:
- `CONTROLLED_VAR`: entities that are associated with experimental variables and that subjected to controlled and targeted perturbations.
- `MEASURED_VAR`: entities that are associated with the variables measured and the object of the measurements.
`BORING`: entities are marked with the tag `BORING` when they are more of descriptive value and not directly associated with causal hypotheses ('boring' is not an ideal choice of word, but it is short...). Typically, these entities are so-called 'reporter' geneproducts, entities used as common baseline across samples, or specify the context of the experiment (cellular system, species, etc...).
### Languages
The text in the dataset is English.
## Dataset Structure
### Data Instances
```json
{
"words": [
".", "Figure", "6", "(", "A", ")", "Cisplatin", "dose", "response", "curves", "of", "(", "i", ")", "MB002", ",", "(", "ii", ")", "Daoy", ",", "and", "(", "iii", ")", "MIC", "in", "the", "absence", "(", "EV", ")", "or", "presence", "of", "SOX9", "by", "Alamar", "blue", ".", "Cells", "were", "pre", "-", "conditioned", "with", "doxycycline", "to", "induce", "expression", "of", "SOX9", "(", "or", "EV", ")", "prior", "to", "treatment", "with", "increasing", "concentrations", "of", "cisplatin", ".", "The", "IC50", "were", "calculated", "following", "5", "(", "MB002", "and", "MIC", ")", "or", "3", "days", "(", "Daoy", ")", "of", "treatment", ".", "Data", "are", "mean", "+", "standard", "deviation", "from", "3", "independent", "repeats", ",", "each", "containing", "5", "technical", "replicates", ".", "(", "B", ")", "Cisplatin", "dose", "response", "curves", "of", "SOX9", "-", "expressing", "(", "i", ")", "Daoy", "and", "(", "ii", ")", "MIC", "in", "the", "absence", "or", "presence", "of", "FBW7\u03b1", ".", "Experiments", "and", "data", "analysis", "were", "performed", "as", "described", "in", "(", "A", ")", "(", "C", ")", "Overall", "survival", "analysis", "of", "mice", "bearing", "Daoy", "or", "Daoy", "-", "expressing", "dox", "-", "inducible", "SOX9", "treated", "with", "cisplatin", ".", "The", "dox", "-", "preconditioned", "cells", "(", "105", "cells", ")", "were", "orthotopically", "xenografted", "to", "Nude", "-", "Foxn1nu", "mice", "and", "left", "for", "1", "week", "to", "prior", "to", "being", "treated", "with", "vehicle", "control", "or", "cisplatin", "(", "2mg", "/", "kg", ")", "intraperitoneally", "for", "every", "other", "day", "for", "a", "total", "of", "6", "doses", ".", "(", "D", ")", "Heat", "map", "of", "the", "row", "-", "wise", "z", "-", "scores", "of", "11", "genes", "associated", "with", "cisplatin", "resistance", "in", "MB002", "expressing", "Sox9", "-", "WT", "or", "Sox9", "-", "T236", "/", "T240A", ".", "Heat", "map", "was", "generated", "using", "the", "GenePattern", "software", ".", "(", "E", ")", "Quantitative", "analysis", "of", "ATP7A", ",", "DUSP2", ",", "and", "TTK", "mRNAs", "in", "MB002", "following", "expression", "of", "SOX9", "-", "WT", "or", "SOX9", "-", "T236", "/", "240A", ".", "Total", "RNA", "were", "collected", "24", "hours", "following", "doxycycline", "treatment", ",", "from", "which", "cDNA", "were", "generated", "for", "qPCR", ".", "Data", "are", "mean", "mRNA", "level", "(", "normalized", "to", "B2M", "transcript", ")", "+", "standard", "deviation", "from", "3", "independent", "experiments", "with", "statistical", "significance", "were", "determined", "by", "Multiple", "comparisons", "2", "-", "way", "ANOVA", "with", "Bonferroni", "'", "s", "post", "-", "test", ".", "(", "F", ")", "Time", "course", "western", "blotting", "of", "HA", "-", "SOX9", ",", "ATP7A", ",", "DUSP2", ",", "ERK1", "/", "2", "pThr202", "/", "Tyr204", "and", "total", "ERK1", "/", "2", "in", "MB002", "cells", "following", "doxycycline", "induction", "of", "either", "EV", ",", "SOX9", "-", "WT", "or", "SOX9", "-", "T236", "/", "240A", ".", "GAPDH", "was", "used", "as", "a", "loading", "control", "."
],
"panel_id": "12345",
"label_ids": {
"entity_types": [
"O", "O", "O", "O", "O", "O", "B-SMALL_MOLECULE", "O", "O", "O", "O", "O", "O", "O", "B-CELL", "O", "O", "O", "O", "B-CELL", "O", "O", "O", "O", "O", "B-CELL", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-GENEPROD", "O", "B-EXP_ASSAY", "I-EXP_ASSAY", "O", "O", "O", "O", "O", "O", "O", "B-SMALL_MOLECULE", "O", "O", "O", "O", "B-GENEPROD", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-SMALL_MOLECULE", "O", "O", "O", "O", "O", "O", "O", "O", "B-CELL", "O", "B-CELL", "O", "O", "O", "O", "O", "B-CELL", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-SMALL_MOLECULE", "O", "O", "O", "O", "B-GENEPROD", "O", "O", "O", "O", "O", "B-CELL", "O", "O", "O", "O", "B-CELL", "O", "O", "O", "O", "O", "O", "B-GENEPROD", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-EXP_ASSAY", "O", "O", "B-ORGANISM", "O", "B-CELL", "O", "B-CELL", "O", "O", "B-SMALL_MOLECULE", "O", "O", "B-GENEPROD", "O", "O", "B-SMALL_MOLECULE", "O", "O", "B-SMALL_MOLECULE", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-ORGANISM", "O", "O", "O", "B-GENEPROD", "B-ORGANISM", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-SMALL_MOLECULE", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-SMALL_MOLECULE", "O", "O", "B-CELL", "O", "B-GENEPROD", "O", "O", "O", "B-GENEPROD", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-GENEPROD", "O", "B-GENEPROD", "O", "O", "B-GENEPROD", "O", "O", "B-CELL", "O", "O", "O", "B-GENEPROD", "O", "O", "O", "B-GENEPROD", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-SMALL_MOLECULE", "O", "O", "O", "O", "O", "O", "O", "O", "B-EXP_ASSAY", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-GENEPROD", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-EXP_ASSAY", "I-EXP_ASSAY", "O", "B-GENEPROD", "O", "B-GENEPROD", "O", "B-GENEPROD", "O", "B-GENEPROD", "O", "B-GENEPROD", "I-GENEPROD", "I-GENEPROD", "O", "O", "O", "O", "O", "B-GENEPROD", "I-GENEPROD", "I-GENEPROD", "O", "B-CELL", "O", "O", "B-SMALL_MOLECULE", "O", "O", "O", "O", "O", "B-GENEPROD", "O", "O", "O", "B-GENEPROD", "O", "O", "O", "O", "O", "B-GENEPROD", "O", "O", "O", "O", "O", "O", "O"
],
"geneprod_roles": [
"O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-CONTROLLED_VAR", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-CONTROLLED_VAR", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-CONTROLLED_VAR", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-CONTROLLED_VAR", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-CONTROLLED_VAR", "O", "O", "O", "B-CONTROLLED_VAR", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-MEASURED_VAR", "O", "B-MEASURED_VAR", "O", "O", "B-MEASURED_VAR", "O", "O", "O", "O", "O", "O", "B-CONTROLLED_VAR", "O", "O", "O", "B-CONTROLLED_VAR", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-MEASURED_VAR", "O", "B-MEASURED_VAR", "O", "B-MEASURED_VAR", "O", "B-MEASURED_VAR", "I-MEASURED_VAR", "I-MEASURED_VAR", "O", "O", "O", "O", "O", "B-MEASURED_VAR", "I-MEASURED_VAR", "I-MEASURED_VAR", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-CONTROLLED_VAR", "O", "O", "O", "B-CONTROLLED_VAR", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O"
],
"boring": [
"O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-BORING", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-BORING", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-BORING", "O", "B-BORING", "O", "O", "B-BORING", "O", "O", "O", "O", "O", "O", "O", "O", "B-BORING", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-BORING", "O", "O", "O", "B-BORING", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-BORING", "O", "O", "B-BORING", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-BORING", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-BORING", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-BORING", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-BORING", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-BORING", "O", "O", "B-BORING", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-BORING", "O", "O", "O", "O", "O", "O", "O"
],
"panel_start": [
"O", "O", "O", "B-PANEL_START", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-PANEL_START", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-PANEL_START", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-PANEL_START", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-PANEL_START", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-PANEL_START", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O"
],
"small_mol_roles": ["O", "O", "O", "O", "O", "O", "B-CONTROLLED_VAR", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-CONTROLLED_VAR", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-CONTROLLED_VAR", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-CONTROLLED_VAR", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-CONTROLLED_VAR", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O"]
}
}
```
### Data Fields
- `words`: `list` of `strings` text tokenized into words.
- `panel_id`: ID of the panel to which the example belongs to in the SourceData database.
- `label_ids`:
- `entity_types`: `list` of `strings` for the IOB2 tags for entity type; possible value in `["O", "I-SMALL_MOLECULE", "B-SMALL_MOLECULE", "I-GENEPROD", "B-GENEPROD", "I-SUBCELLULAR", "B-SUBCELLULAR", "I-CELL", "B-CELL", "I-TISSUE", "B-TISSUE", "I-ORGANISM", "B-ORGANISM", "I-EXP_ASSAY", "B-EXP_ASSAY"]`
- `geneprod_roles`: `list` of `strings` for the IOB2 tags for experimental roles; values in `["O", "I-CONTROLLED_VAR", "B-CONTROLLED_VAR", "I-MEASURED_VAR", "B-MEASURED_VAR"]`
- `boring`: `list` of `strings` for IOB2 tags for entities unrelated to causal design; values in `["O", "I-BORING", "B-BORING"]`
- `panel_start`: `list` of `strings` for IOB2 tags `["O", "B-PANEL_START"]`
- `small_mol_roles`: `list` of `strings` for IOB2 tags showing whether the entity is the variable being measured or the control variable `["O", "B-CONTROLLED_VAR", "I-CONTROLLED_VAR", "B-MEASURED_VAR", "I-MEASURED_VAR",]`
### Data Splits
- train:
- features: ['words', 'labels', 'tag_mask', 'panel_id'],
- num_rows: 50_198
- validation:
- features: ['words', 'labels', 'tag_mask', 'panel_id'],
- num_rows: 5_946
- test:
- features: ['words', 'labels', 'tag_mask', 'panel_id'],
- num_rows: 6_222
## Dataset Creation
### Curation Rationale
The dataset was built to train models for the automatic extraction of a knowledge graph based from the scientific literature. The dataset can be used to train models for text segmentation, named entity recognition and semantic role labeling.
### Source Data
#### Initial Data Collection and Normalization
Figure legends were annotated according to the SourceData framework described in Liechti et al 2017 (Nature Methods, 2017, https://doi.org/10.1038/nmeth.4471). The curation tool at https://curation.sourcedata.io was used to segment figure legends into panel legends, tag enities, assign experiemental roles and normalize with standard identifiers (not available in this dataset). The source data was downloaded from the SourceData API (https://api.sourcedata.io) on 21 Jan 2021.
#### Who are the source language producers?
The examples are extracted from the figure legends from scientific papers in cell and molecular biology.
### Annotations
#### Annotation process
The annotations were produced manually with expert curators from the SourceData project (https://sourcedata.embo.org)
#### Who are the annotators?
Curators of the SourceData project.
### Personal and Sensitive Information
None known.
## Considerations for Using the Data
### Social Impact of Dataset
Not applicable.
### Discussion of Biases
The examples are heavily biased towards cell and molecular biology and are enriched in examples from papers published in EMBO Press journals (https://embopress.org)
The annotation of diseases has been added recently to the dataset. Although they appear, the number is very low and they are not consistently tagged through the entire dataset.
We recommend to use the diseases by filtering the examples that contain them.
### Other Known Limitations
[More Information Needed]
## Additional Information
### Dataset Curators
Thomas Lemberger, EMBO.
Jorge Abreu Vicente, EMBO
### Licensing Information
CC BY 4.0
### Citation Information
We are currently working on a paper to present the dataset. It is expected to be ready by 2023 spring. In the meantime, the following paper should be cited.
```latex
@article {Liechti2017,
author = {Liechti, Robin and George, Nancy and Götz, Lou and El-Gebali, Sara and Chasapi, Anastasia and Crespo, Isaac and Xenarios, Ioannis and Lemberger, Thomas},
title = {SourceData - a semantic platform for curating and searching figures},
year = {2017},
volume = {14},
number = {11},
doi = {10.1038/nmeth.4471},
URL = {https://doi.org/10.1038/nmeth.4471},
eprint = {https://www.biorxiv.org/content/early/2016/06/20/058529.full.pdf},
journal = {Nature Methods}
}
```
### Contributions
Thanks to [@tlemberger](https://github.com/tlemberger>) and [@drAbreu](https://github.com/drAbreu>) for adding this dataset.
|
false |
# Dataset Card for DigitalUmuganda/common-voice-kinyarwanda-text-dataset
|
false |
# DCASE 2022 Task 3 Data sets: STARSS22 Dataset + Synthetic SELD mixtures
[Audio Research Group / Tampere University](https://webpages.tuni.fi/arg/)
[Creative AI Lab/ SONY R&D Center](https://www.sony.com/en/SonyInfo/research/research-areas/audio-acoustics/)
## Important
**This is a copy from the Zenodo Original one**
AUTHORS
**Tampere University**
- Archontis Politis ([contact](mailto:archontis.politis@tuni.fi), [profile](https://scholar.google.fi/citations?user=DuCqB3sAAAAJ&hl=en))
- Parthasaarathy Sudarsanam([contact](mailto:parthasaarathy.ariyakulamsudarsanam@tuni.fi), [profile](https://scholar.google.com/citations?user=yxZ1qAIAAAAJ&hl=en))
- Sharath Adavanne ([contact](mailto:sharath.adavanne@tuni.fi), [profile](https://www.aane.in))
- Daniel Krause ([contact](mailto:daniel.krause@tuni.fi), [profile](https://scholar.google.com/citations?user=pSLng-8AAAAJ&hl=en))
- Tuomas Virtanen ([contact](mailto:tuomas.virtanen@tuni.fi), [profile](https://homepages.tuni.fi/tuomas.virtanen/))
**SONY**
- Yuki Mitsufuji ([contact](mailto:yuhki.mitsufuji@sony.com), [profile](https://scholar.google.com/citations?user=GMytI10AAAAJ))
- Kazuki Shimada ([contact](mailto:kazuki.shimada@sony.com), [profile](https://scholar.google.com/citations?user=-t9IslAAAAAJ&hl=en))
- Naoya Takahashi ([profile](https://scholar.google.com/citations?user=JbtYJMoAAAAJ))
- Yuichiro Koyama
- Shusuke Takahashi
# Description
The **Sony-TAu Realistic Spatial Soundscapes 2022 (STARSS22)** dataset contains multichannel recordings of sound scenes in various rooms and environments, together with temporal and spatial annotations of prominent events belonging to a set of target classes. The dataset is collected in two different countries, in Tampere, Finland by the Audio Researh Group (ARG) of **Tampere University (TAU)**, and in Tokyo, Japan by **SONY**, using a similar setup and annotation procedure. The dataset is delivered in two 4-channel spatial recording formats, a microphone array one (**MIC**), and first-order Ambisonics one (**FOA**). These recordings serve as the development dataset for the [DCASE 2022 Sound Event Localization and Detection Task](https://dcase.community/challenge2022/task-sound-event-localization-and-detection) of the [DCASE 2022 Challenge](https://dcase.community/challenge2022/).
Contrary to the three previous datasets of synthetic spatial sound scenes of TAU Spatial Sound Events 2019 ([development](10.5281/zenodo.2599196)/[evaluation](10.5281/zenodo.3377088)), [TAU-NIGENS Spatial Sound Events 2020](https://doi.org/10.5281/zenodo.4064792), and [TAU-NIGENS Spatial Sound Events 2021](10.5281/zenodo.5476980
) associated with the previous iterations of the DCASE Challenge, the STARSS22 dataset contains recordings of real sound scenes and hence it avoids some of the pitfalls of synthetic generation of scenes. Some such key properties are:
- annotations are based on a combination of human annotators for sound event activity and optical tracking for spatial positions
- the annotated target event classes are determined by the composition of the real scenes
- the density, polyphony, occurences and co-occurences of events and sound classes is not random, and it follows actions and interactions of participants in the real scenes
The recordings were collected between September 2021 and February 2022. Collection of data from the TAU side has received funding from Google.
# Aim
The dataset is suitable for training and evaluation of machine-listening models for sound event detection (SED), general sound source localization with diverse sounds or signal-of-interest localization, and joint sound-event-localization-and-detection (SELD). Additionally, the dataset can be used for evaluation of signal processing methods that do not necessarily rely on training, such as acoustic source localization methods and multiple-source acoustic tracking. The dataset allows evaluation of the performance and robustness of the aforementioned applications for diverse types of sounds, and under diverse acoustic conditions.
# Recording procedure
The sound scene recordings were captured with a high-channel-count spherical microphone array ([Eigenmike em32 by mh Acoustics](https://mhacoustics.com/products)), simultaneously with a 360° video recording spatially aligned with the spherical array recording ([Ricoh Theta V](https://theta360.com/en/about/theta/v.html)). Additionally, the main sound sources of interest were equipped with tracking markers, which are tracked throughout the recording with an [Optitrack Flex 13](https://optitrack.com/cameras/flex-13/) system arranged around each scene. All scenes were based on human actors performing some actions, interacting between them and with the objects in the scene, and were by design dynamic. Since the actors were producing most of the sounds in the scene (but not all), they were additionally equipped with [DPA Wireless Go II](https://rode.com/microphones/wireless/wirelessgoii) microphones, providing close-miked recordings of the main events. Recording would start and stop according to a scene being acted, usually lasting between 1~5mins. Recording would start in all microphones and tracking devices before the beginning of the scene, and would stop right after. A clapper sound would initiate the acting and it would serve as a reference signal for synchronization between the em32 recording, the Ricoh Theta V video, the DPA wireless microphone recordings, and the Optitrack tracker data. Synchronized clips of all of them would be cropped and stored in the end of each recording session.
# Annotation procedure
By combining information from the wireless microphones, the optical tracking data, and the 360° videos, spatiotemporal annotations were extracted semi-automatically, and validated manually. More specifically, the actors were tracked all through each recording session wearing headbands with markers, and the spatial positions of other human-related sources, such as mouth, hands, or footsteps were geometrically extrapolated from those head coordinates. Additional trackers were mounted on other sources of interest (e.g. vacuum cleaner, guitar, water tap, cupboard, door handle, a.o.). Each actor had a wireless microphone mounted on their lapel, providing a clear recording of all sound events produced by that actor, and/or any independent sources closer to that actor than the rest. The temporal annotation was based primarily on those close-miked recordings. The annotators would annotate the sound event activity and label their class during the recording by listening those close-miked signals. Events that were not audible in the overall scene recording of the em32 were not annotated, even if they were audible in the lapel recordings. In ambiguous cases, the annotators could rely on the 360° video to associate an event with a certain actor or source. The final sound event temporal annotations were associated with the tracking data through the class of each sound event and the actor that produced them. All tracked Cartesian coordinates delivered by the tracker were converted to directions-of-arrival (DOAs) with respect to the coordinates of the Eigenmike. Finally, the final class, temporal, and spatial annotations were combined and converted to the challenge format. Validation of the annotations was done by observing videos of the activities of each class visualized as markers positioned at their respective DOAs on the 360° video plane, overlapped with the 360° from the Ricoh Theta V.
# Recording formats
The array response of the two recording formats can be considered known. The following theoretical spatial responses (steering vectors) modeling the two formats describe the directional response of each channel to a source incident from direction-of-arrival (DOA) given by azimuth angle $\phi$ and elevation angle $\theta$.
**For the first-order ambisonics (FOA):**
\begin{eqnarray}
H_1(\phi, \theta, f) &=& 1 \\
H_2(\phi, \theta, f) &=& \sin(\phi) * \cos(\theta) \\
H_3(\phi, \theta, f) &=& \sin(\theta) \\
H_4(\phi, \theta, f) &=& \cos(\phi) * \cos(\theta)
\end{eqnarray}
The (FOA) format is obtained by converting the 32-channel microphone array signals by means of encoding filters based on anechoic measurements of the Eigenmike array response. Note that in the formulas above the encoding format is assumed frequency-independent, something that holds true up to around 9kHz with the specific microphone array, while the actual encoded responses start to deviate gradually at higher frequencies from the ideal ones provided above.
**For the tetrahedral microphone array (MIC):**
The four microphone have the following positions, in spherical coordinates $(\phi, \theta, r)$:
\begin{eqnarray}
M1: &\quad(&45^\circ, &&35^\circ, &4.2\mathrm{cm})\nonumber\\
M2: &\quad(&-45^\circ, &-&35^\circ, &4.2\mathrm{cm})\nonumber\\
M3: &\quad(&135^\circ, &-&35^\circ, &4.2\mathrm{cm})\nonumber\\
M4: &\quad(&-135^\circ, &&35^\circ, &4.2\mathrm{cm})\nonumber
\end{eqnarray}
Since the microphones are mounted on an acoustically-hard spherical baffle, an analytical expression for the directional array response is given by the expansion:
\begin{equation}
H_m(\phi_m, \theta_m, \phi, \theta, \omega) = \frac{1}{(\omega R/c)^2}\sum_{n=0}^{30} \frac{i^{n-1}}{h_n'^{(2)}(\omega R/c)}(2n+1)P_n(\cos(\gamma_m))
\end{equation}
where $m$ is the channel number, $(\phi_m, \theta_m)$ are the specific microphone's azimuth and elevation position, $\omega = 2\pi f$ is the angular frequency, $R = 0.042$m is the array radius, $c = 343$m/s is the speed of sound, $\cos(\gamma_m)$ is the cosine angle between the microphone and the DOA, and $P_n$ is the unnormalized Legendre polynomial of degree $n$, and $h_n'^{(2)}$ is the derivative with respect to the argument of a spherical Hankel function of the second kind. The expansion is limited to 30 terms which provides negligible modeling error up to 20kHz. Example routines that can generate directional frequency and impulse array responses based on the above formula can be found [here](https://github.com/polarch/Array-Response-Simulator).
# Dataset specifications
The specifications of the dataset can be summarized in the following:
- 70 recording clips of 30 sec ~ 5 min durations, with a total time of ~2hrs, contributed by SONY (development dataset).
- 51 recording clips of 1 min ~ 5 min durations, with a total time of ~3hrs, contributed by TAU (development dataset).
- A training-test split is provided for reporting results using the development dataset.
- 40 recordings contributed by SONY for the training split, captured in 2 rooms (dev-train-sony).
- 30 recordings contributed by SONY for the testing split, captured in 2 rooms (dev-test-sony).
- 27 recordings contributed by TAU for the training split, captured in 4 rooms (dev-train-tau).
- 24 recordings contributed by TAU for the testing split, captured in 3 rooms (dev-test-tau).
- A total of 11 unique rooms captured in the recordings, 4 from SONY and 7 from TAU (development set).
- Sampling rate 24kHz.
- Two 4-channel 3-dimensional recording formats: first-order Ambisonics (FOA) and tetrahedral microphone array (MIC).
- Recordings are taken in two different countries and two different sites.
- Each recording clip is part of a recording session happening in a unique room.
- Groups of participants, sound making props, and scene scenarios are unique for each session (with a few exceptions).
- To achieve good variability and efficiency in the data, in terms of presence, density, movement, and/or spatial distribution of the sounds events, the scenes are loosely scripted.
- 13 target classes are identified in the recordings and strongly annotated by humans.
- Spatial annotations for those active events are captured by an optical tracking system.
- Sound events out of the target classes are considered as interference.
# Sound event classes
13 target sound event classes were annotated. The classes follow loosely the [Audioset ontology](https://research.google.com/audioset/ontology/index.html).
0. Female speech, woman speaking
1. Male speech, man speaking
2. Clapping
3. Telephone
4. Laughter
5. Domestic sounds
6. Walk, footsteps
7. Door, open or close
8. Music
9. Musical instrument
10. Water tap, faucet
11. Bell
12. Knock
The content of some of these classes corresponds to events of a limited range of Audioset-related subclasses. These are detailed here as additional information on the diversity of those sound events:
- Telephone
- Mostly traditional _Telephone Bell Ringing_ and _Ringtone_ sounds, without musical ringtones.
- Domestic sounds
- Sounds of _Vacuum cleaner_
- Sounds of water boiler, closer to _Boiling_
- Sounds of air circulator, closer to _Mechanical fan_
- Door, open or close
- Combination of _Door_ and _Cupboard open or close_
- Music
- _Background music_ and _Pop music_ played by a loudspeaker in the room.
- Musical Instrument
- Acoustic guitar
- Marimba, xylophone
- Cowbell
- Piano
- Rattle (instrument)
- Bell
- Combination of sounds from hotel bell and glass bell, closer to _Bicycle bell_ and single _Chime_.
Some additional notes:
- The speech classes contain speech in a few different languages.
- There are occasionally localized sound events that are not annotated and are considered as interferers, with examples such as _computer keyboard_, _shuffling cards_, _dishes, pots, and pans_.
- There is natural background noise (e.g. HVAC noise) in all recordings, at very low levels in some and at quite high levels in others. Such mostly diffuse background noise should be distinct from other noisy target sources (e.g. vacuum cleaner, mechanical fan) since these are clearly spatially localized.
# Naming Convention (Development dataset)
The recordings in the development dataset follow the naming convention:
fold[fold number]_room[room number]_mix[recording number per room].wav
The fold number at the moment is used only to distinguish between the training and testing split. The room information is provided for the user of the dataset to potentially help understand the performance of their method with respect to different conditions.
# Reference labels and directions-of-arrival
For each recording in the development dataset, the labels and DoAs are provided in a plain text CSV file of the same filename as the recording, in the following format:
[frame number (int)], [active class index (int)], [source number index (int)], [azimuth (int)], [elevation (int)]
Frame, class, and source enumeration begins at 0. Frames correspond to a temporal resolution of 100msec. Azimuth and elevation angles are given in degrees, rounded to the closest integer value, with azimuth and elevation being zero at the front, azimuth $\phi \in [-180^{\circ}, 180^{\circ}]$, and elevation $\theta \in [-90^{\circ}, 90^{\circ}]$. Note that the azimuth angle is increasing counter-clockwise ($\phi = 90^{\circ}$ at the left).
The source index is a unique integer for each source in the scene, and it is provided only as additional information. Note that each unique actor gets assigned one such identifier, but not individual events produced by the same actor; e.g. a _clapping_ event and a _laughter_ event produced by the same person have the same identifier. Independent sources that are not actors (e.g. a loudspeaker playing music in the room) get a 0 identifier. Note that source identifier information is only included in the development metadata and is not required to be provided by the participants in their results.
Overlapping sound events are indicated with duplicate frame numbers, and can belong to a different or the same class. An example sequence could be as:
10, 1, 1, -50, 30
11, 1, 1, -50, 30
11, 1, 2, 10, -20
12, 1, 2, 10, -20
13, 1, 2, 10, -20
13, 8, 0, -40, 0
which describes that in frame 10-11, an event of class _male speech_ (_class 1_) belonging to one actor (_source 1_) is active at direction (-50°,30°). However, at frame 11 a second instance of the same class appears simultaneously at a different direction (10°,-20°) belonging to another actor (_source 2_), while at frame 13 an additional event of class _music_ (_class 8_) appears belonging to a non-actor source (_source 0_). Frames that contain no sound events are not included in the sequence.
# Task setup
The dataset is associated with the [DCASE 2022 Challenge](http://dcase.community/challenge2022/). To have consistent reporting of results between participants on the development set a pre-defined training-testing split is provided. To compare against the challenge baseline and with other participants during the development stage, models should be trained on the training split only, and results should be reported on the testing split only.
**Note that even though there are two origins of the data, SONY and TAU, the challenge task considers the dataset as a single entity. Hence models should not be trained separately for each of the two origins, and tested individually on recordings of each of them. Instead, the recordings of the individual training splits (_dev-test-sony_, _dev_test_tau_) and testing splits (_dev-test-sony_, _dev_test_tau_) should be combined (_dev_train_, _dev_test_) and the models should be trained and evaluated in the respective combined splits.**
The evaluation part of the dataset will be published here as a new dataset version, a few weeks before the final challenge submission deadline. The additional evaluation files consist of only audio recordings without any metadata/labels. Participants can decide the training procedure, i.e. the amount of training and validation files in the development dataset, the number of ensemble models etc., and submit the results of the SELD performance on the evaluation dataset.
# File structure
```
dataset root
│ README.md this file, markdown-format
| LICENSE the license file
│
└───foa_dev Ambisonic format, 24kHz, four channels
| | dev-train-sony to be used for training when reporting development set results (SONY recordings)
│ │ | fold3_room21_mix001.wav
│ │ | fold3_room21_mix002.wav
│ │ | ...
│ │ | fold3_room22_mix001.wav
│ │ | fold3_room22_mix002.wav
│ | │ ...
| | dev-test-sony to be used for testing when reporting development set results (SONY recordings)
│ │ | fold4_room23_mix001.wav
│ │ | fold4_room23_mix002.wav
│ │ | ...
│ │ | fold4_room24_mix001.wav
│ │ | fold4_room24_mix002.wav
│ │ | ...
| | dev-train-tau to be used for training when reporting development set results (TAU recordings)
│ │ | fold3_room4_mix001.wav
│ │ | fold3_room4_mix002.wav
│ │ | ...
│ │ | fold3_room6_mix001.wav
│ │ | fold3_room6_mix002.wav
│ | │ ...
│ │ | fold3_room7_mix001.wav
│ │ | fold3_room7_mix002.wav
│ | │ ...
│ │ | fold3_room9_mix001.wav
│ │ | fold3_room9_mix002.wav
│ | │ ...
| | dev-test-tau to be used for testing when reporting development set results (TAU recordings)
│ │ | fold4_room2_mix001.wav
│ │ | fold4_room2_mix002.wav
│ │ | ...
│ │ | fold4_room8_mix001.wav
│ │ | fold4_room8_mix002.wav
│ │ | ...
│ │ | fold4_room10_mix001.wav
│ │ | fold4_room10_mix002.wav
│ │ | ...
│
└───mic_dev Microphone array format, 24kHz, four channels
| | dev-train-sony to be used for training when reporting development set results (SONY recordings)
│ │ | fold3_room21_mix001.wav
│ │ | fold3_room21_mix002.wav
│ │ | ...
│ │ | fold3_room22_mix001.wav
│ │ | fold3_room22_mix002.wav
│ | │ ...
| | dev-test-sony to be used for testing when reporting development set results (SONY recordings)
│ │ | fold4_room23_mix001.wav
│ │ | fold4_room23_mix002.wav
│ │ | ...
│ │ | fold4_room24_mix001.wav
│ │ | fold4_room24_mix002.wav
│ │ | ...
| | dev-train-tau to be used for training when reporting development set results (TAU recordings)
│ │ | fold3_room4_mix001.wav
│ │ | fold3_room4_mix002.wav
│ │ | ...
│ │ | fold3_room6_mix001.wav
│ │ | fold3_room6_mix002.wav
│ | │ ...
│ │ | fold3_room7_mix001.wav
│ │ | fold3_room7_mix002.wav
│ | │ ...
│ │ | fold3_room9_mix001.wav
│ │ | fold3_room9_mix002.wav
│ | │ ...
| | dev-test-tau to be used for testing when reporting development set results (TAU recordings)
│ │ | fold4_room2_mix001.wav
│ │ | fold4_room2_mix002.wav
│ │ | ...
│ │ | fold4_room8_mix001.wav
│ │ | fold4_room8_mix002.wav
│ │ | ...
│ │ | fold4_room10_mix001.wav
│ │ | fold4_room10_mix002.wav
│ │ | ...
│
└───metadata_dev `csv` format, 600 files
| | dev-train-sony to be used for training when reporting development set results (SONY recordings)
│ │ | fold3_room21_mix001.csv
│ │ | fold3_room21_mix002.csv
│ │ | ...
│ │ | fold3_room22_mix001.csv
│ │ | fold3_room22_mix002.csv
│ | │ ...
| | dev-test-sony to be used for testing when reporting development set results (SONY recordings)
│ │ | fold4_room23_mix001.csv
│ │ | fold4_room23_mix002.csv
│ │ | ...
│ │ | fold4_room24_mix001.csv
│ │ | fold4_room24_mix002.csv
│ │ | ...
| | dev-train-tau to be used for training when reporting development set results (TAU recordings)
│ │ | fold3_room4_mix001.csv
│ │ | fold3_room4_mix002.csv
│ │ | ...
│ │ | fold3_room6_mix001.csv
│ │ | fold3_room6_mix002.csv
│ | │ ...
│ │ | fold3_room7_mix001.csv
│ │ | fold3_room7_mix002.csv
│ | │ ...
│ │ | fold3_room9_mix001.csv
│ │ | fold3_room9_mix002.csv
│ | │ ...
| | dev-test-tau to be used for testing when reporting development set results (TAU recordings)
│ │ | fold4_room2_mix001.csv
│ │ | fold4_room2_mix002.csv
│ │ | ...
│ │ | fold4_room8_mix001.csv
│ │ | fold4_room8_mix002.csv
│ │ | ...
│ │ | fold4_room10_mix001.csv
│ │ | fold4_room10_mix002.csv
│ │ | ...
```
# Download
git clone
# Example application
An implementation of a trainable model of a convolutional recurrent neural network, performing joint SELD, trained and evaluated with this dataset is provided [here](https://github.com/sharathadavanne/seld-dcase2022). This implementation will serve as the baseline method in the DCASE 2022 Sound Event Localization and Detection Task.
# License
This datast is licensed under the [MIT](https://opensource.org/licenses/MIT) license.
|
false |
# Dataset Card for "lmqg/qg_korquad"
## Dataset Description
- **Repository:** [https://github.com/asahi417/lm-question-generation](https://github.com/asahi417/lm-question-generation)
- **Paper:** [https://arxiv.org/abs/2210.03992](https://arxiv.org/abs/2210.03992)
- **Point of Contact:** [Asahi Ushio](http://asahiushio.com/)
### Dataset Summary
This is a subset of [QG-Bench](https://github.com/asahi417/lm-question-generation/blob/master/QG_BENCH.md#datasets), a unified question generation benchmark proposed in
["Generative Language Models for Paragraph-Level Question Generation: A Unified Benchmark and Evaluation, EMNLP 2022 main conference"](https://arxiv.org/abs/2210.03992).
This is a modified version of [KorQuAD](https://huggingface.co/datasets/squad_kor_v1) for question generation (QG) task.
Since the original dataset only contains training/validation set, we manually sample test set from training set, which
has no overlap in terms of the paragraph with the training set.
### Supported Tasks and Leaderboards
* `question-generation`: The dataset is assumed to be used to train a model for question generation.
Success on this task is typically measured by achieving a high BLEU4/METEOR/ROUGE-L/BERTScore/MoverScore (see our paper for more in detail).
### Languages
Korean (ko)
## Dataset Structure
An example of 'train' looks as follows.
```
{
"question": "함수해석학이 주목하는 탐구는?",
"paragraph": "변화에 대한 이해와 묘사는 자연과학에 있어서 일반적인 주제이며, 미적분학은 변화를 탐구하는 강력한 도구로서 발전되었다. 함수는 변화하는 양을 묘사함에 있어서 중추적인 개념으로써 떠오르게 된다. 실수와 실변수로 구성된 함수의 엄밀한 탐구가 실해석학이라는 분야로 알려지게 되었고, 복소수에 대한 이와 같은 탐구분야는 복소해석학이라고 한다. 함수해석학은 함수의 공간(특히 무한차원)의 탐구에 주목한다. 함수해석학의 많은 응용분야 중 하나가 양자역학이다. 많은 문제들이 자연스럽게 양과 그 양의 변화율의 관계로 귀착되고, 이러한 문제들이 미분방정식으로 다루어진다. 자연의 많은 현상들이 동역학계로 기술될 수 있다. 혼돈 이론은 이러한 예측 불가능한 현상을 탐구하는 데 상당한 기여를 한다.",
"answer": "함수의 공간(특히 무한차원)의 탐구",
"sentence": "함수해석학은 함수의 공간(특히 무한차원)의 탐구 에 주목한다.",
"paragraph_sentence": '변화에 대한 이해와 묘사는 자연과학에 있어서 일반적인 주제이며, 미적분학은 변화를 탐구하는 강력한 도구로서 발전되었다. 함수는 변화하는 양을 묘사함에 있어서 중추적인 개념으로써 떠오르게 된다. 실수와 실변수로 구성된 함수의 엄밀한 탐구가 실해석학이라는 분야로 알려지게 되었고, 복소수에 대한 이와 같은 탐구 분야는 복소해석학이라고 한다. <hl> 함수해석학은 함수의 공간(특히 무한차원)의 탐구 에 주목한다. <hl> 함수해석학의 많은 응용분야 중 하나가 양자역학이다. 많은 문제들이 자연스럽게 양과 그 양의 변화율의 관계로 귀착되고, 이러한 문제들이 미분방정식으로 다루어진다. 자연의 많은 현상들이 동역학계로 기술될 수 있다. 혼돈 이론은 이러한 예측 불가능한 현상을 탐구하는 데 상당한 기여를 한다.',
"paragraph_answer": '변화에 대한 이해와 묘사는 자연과학에 있어서 일반적인 주제이며, 미적분학은 변화를 탐구하는 강력한 도구로서 발전되었다. 함수는 변화하는 양을 묘사함에 있어서 중추적인 개념으로써 떠오르게 된다. 실수와 실변수로 구성된 함수의 엄밀한 탐구가 실해석학이라는 분야로 알려지게 되었고, 복소수에 대한 이와 같은 탐구 분야는 복소해석학이라고 한다. 함수해석학은 <hl> 함수의 공간(특히 무한차원)의 탐구 <hl>에 주목한다. 함수해석학의 많은 응용분야 중 하나가 양자역학이다. 많은 문제들이 자연스럽게 양과 그 양의 변화율의 관계로 귀착되고, 이러한 문제들이 미분방정식으로 다루어진다. 자연의 많은 현상들이 동역학계로 기술될 수 있다. 혼돈 이론은 이러한 예측 불가능한 현상을 탐구하는 데 상당한 기여를 한다.',
"sentence_answer": "함수해석학은 <hl> 함수의 공간(특히 무한차원)의 탐구 <hl> 에 주목한다."
}
```
The data fields are the same among all splits.
- `question`: a `string` feature.
- `paragraph`: a `string` feature.
- `answer`: a `string` feature.
- `sentence`: a `string` feature.
- `paragraph_answer`: a `string` feature, which is same as the paragraph but the answer is highlighted by a special token `<hl>`.
- `paragraph_sentence`: a `string` feature, which is same as the paragraph but a sentence containing the answer is highlighted by a special token `<hl>`.
- `sentence_answer`: a `string` feature, which is same as the sentence but the answer is highlighted by a special token `<hl>`.
Each of `paragraph_answer`, `paragraph_sentence`, and `sentence_answer` feature is assumed to be used to train a question generation model,
but with different information. The `paragraph_answer` and `sentence_answer` features are for answer-aware question generation and
`paragraph_sentence` feature is for sentence-aware question generation.
## Data Splits
|train|validation|test |
|----:|---------:|----:|
|54556| 5766 |5766 |
## Citation Information
```
@inproceedings{ushio-etal-2022-generative,
title = "{G}enerative {L}anguage {M}odels for {P}aragraph-{L}evel {Q}uestion {G}eneration: {A} {U}nified {B}enchmark and {E}valuation",
author = "Ushio, Asahi and
Alva-Manchego, Fernando and
Camacho-Collados, Jose",
booktitle = "Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing",
month = dec,
year = "2022",
address = "Abu Dhabi, U.A.E.",
publisher = "Association for Computational Linguistics",
}
``` |
false |
# Dataset Card for BEIR Benchmark
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Homepage:** https://github.com/UKPLab/beir
- **Repository:** https://github.com/UKPLab/beir
- **Paper:** https://openreview.net/forum?id=wCu6T5xFjeJ
- **Leaderboard:** https://docs.google.com/spreadsheets/d/1L8aACyPaXrL8iEelJLGqlMqXKPX2oSP_R10pZoy77Ns
- **Point of Contact:** nandan.thakur@uwaterloo.ca
### Dataset Summary
BEIR is a heterogeneous benchmark that has been built from 18 diverse datasets representing 9 information retrieval tasks:
- Fact-checking: [FEVER](http://fever.ai), [Climate-FEVER](http://climatefever.ai), [SciFact](https://github.com/allenai/scifact)
- Question-Answering: [NQ](https://ai.google.com/research/NaturalQuestions), [HotpotQA](https://hotpotqa.github.io), [FiQA-2018](https://sites.google.com/view/fiqa/)
- Bio-Medical IR: [TREC-COVID](https://ir.nist.gov/covidSubmit/index.html), [BioASQ](http://bioasq.org), [NFCorpus](https://www.cl.uni-heidelberg.de/statnlpgroup/nfcorpus/)
- News Retrieval: [TREC-NEWS](https://trec.nist.gov/data/news2019.html), [Robust04](https://trec.nist.gov/data/robust/04.guidelines.html)
- Argument Retrieval: [Touche-2020](https://webis.de/events/touche-20/shared-task-1.html), [ArguAna](tp://argumentation.bplaced.net/arguana/data)
- Duplicate Question Retrieval: [Quora](https://www.quora.com/q/quoradata/First-Quora-Dataset-Release-Question-Pairs), [CqaDupstack](http://nlp.cis.unimelb.edu.au/resources/cqadupstack/)
- Citation-Prediction: [SCIDOCS](https://allenai.org/data/scidocs)
- Tweet Retrieval: [Signal-1M](https://research.signal-ai.com/datasets/signal1m-tweetir.html)
- Entity Retrieval: [DBPedia](https://github.com/iai-group/DBpedia-Entity/)
All these datasets have been preprocessed and can be used for your experiments.
```python
```
### Supported Tasks and Leaderboards
The dataset supports a leaderboard that evaluates models against task-specific metrics such as F1 or EM, as well as their ability to retrieve supporting information from Wikipedia.
The current best performing models can be found [here](https://eval.ai/web/challenges/challenge-page/689/leaderboard/).
### Languages
All tasks are in English (`en`).
## Dataset Structure
All BEIR datasets must contain a corpus, queries and qrels (relevance judgments file). They must be in the following format:
- `corpus` file: a `.jsonl` file (jsonlines) that contains a list of dictionaries, each with three fields `_id` with unique document identifier, `title` with document title (optional) and `text` with document paragraph or passage. For example: `{"_id": "doc1", "title": "Albert Einstein", "text": "Albert Einstein was a German-born...."}`
- `queries` file: a `.jsonl` file (jsonlines) that contains a list of dictionaries, each with two fields `_id` with unique query identifier and `text` with query text. For example: `{"_id": "q1", "text": "Who developed the mass-energy equivalence formula?"}`
- `qrels` file: a `.tsv` file (tab-seperated) that contains three columns, i.e. the `query-id`, `corpus-id` and `score` in this order. Keep 1st row as header. For example: `q1 doc1 1`
### Data Instances
A high level example of any beir dataset:
```python
corpus = {
"doc1" : {
"title": "Albert Einstein",
"text": "Albert Einstein was a German-born theoretical physicist. who developed the theory of relativity, \
one of the two pillars of modern physics (alongside quantum mechanics). His work is also known for \
its influence on the philosophy of science. He is best known to the general public for his mass–energy \
equivalence formula E = mc2, which has been dubbed 'the world's most famous equation'. He received the 1921 \
Nobel Prize in Physics 'for his services to theoretical physics, and especially for his discovery of the law \
of the photoelectric effect', a pivotal step in the development of quantum theory."
},
"doc2" : {
"title": "", # Keep title an empty string if not present
"text": "Wheat beer is a top-fermented beer which is brewed with a large proportion of wheat relative to the amount of \
malted barley. The two main varieties are German Weißbier and Belgian witbier; other types include Lambic (made\
with wild yeast), Berliner Weisse (a cloudy, sour beer), and Gose (a sour, salty beer)."
},
}
queries = {
"q1" : "Who developed the mass-energy equivalence formula?",
"q2" : "Which beer is brewed with a large proportion of wheat?"
}
qrels = {
"q1" : {"doc1": 1},
"q2" : {"doc2": 1},
}
```
### Data Fields
Examples from all configurations have the following features:
### Corpus
- `corpus`: a `dict` feature representing the document title and passage text, made up of:
- `_id`: a `string` feature representing the unique document id
- `title`: a `string` feature, denoting the title of the document.
- `text`: a `string` feature, denoting the text of the document.
### Queries
- `queries`: a `dict` feature representing the query, made up of:
- `_id`: a `string` feature representing the unique query id
- `text`: a `string` feature, denoting the text of the query.
### Qrels
- `qrels`: a `dict` feature representing the query document relevance judgements, made up of:
- `_id`: a `string` feature representing the query id
- `_id`: a `string` feature, denoting the document id.
- `score`: a `int32` feature, denoting the relevance judgement between query and document.
### Data Splits
| Dataset | Website| BEIR-Name | Type | Queries | Corpus | Rel D/Q | Down-load | md5 |
| -------- | -----| ---------| --------- | ----------- | ---------| ---------| :----------: | :------:|
| MSMARCO | [Homepage](https://microsoft.github.io/msmarco/)| ``msmarco`` | ``train``<br>``dev``<br>``test``| 6,980 | 8.84M | 1.1 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/msmarco.zip) | ``444067daf65d982533ea17ebd59501e4`` |
| TREC-COVID | [Homepage](https://ir.nist.gov/covidSubmit/index.html)| ``trec-covid``| ``test``| 50| 171K| 493.5 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/trec-covid.zip) | ``ce62140cb23feb9becf6270d0d1fe6d1`` |
| NFCorpus | [Homepage](https://www.cl.uni-heidelberg.de/statnlpgroup/nfcorpus/) | ``nfcorpus`` | ``train``<br>``dev``<br>``test``| 323 | 3.6K | 38.2 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/nfcorpus.zip) | ``a89dba18a62ef92f7d323ec890a0d38d`` |
| BioASQ | [Homepage](http://bioasq.org) | ``bioasq``| ``train``<br>``test`` | 500 | 14.91M | 8.05 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#2-bioasq) |
| NQ | [Homepage](https://ai.google.com/research/NaturalQuestions) | ``nq``| ``train``<br>``test``| 3,452 | 2.68M | 1.2 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/nq.zip) | ``d4d3d2e48787a744b6f6e691ff534307`` |
| HotpotQA | [Homepage](https://hotpotqa.github.io) | ``hotpotqa``| ``train``<br>``dev``<br>``test``| 7,405 | 5.23M | 2.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/hotpotqa.zip) | ``f412724f78b0d91183a0e86805e16114`` |
| FiQA-2018 | [Homepage](https://sites.google.com/view/fiqa/) | ``fiqa`` | ``train``<br>``dev``<br>``test``| 648 | 57K | 2.6 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/fiqa.zip) | ``17918ed23cd04fb15047f73e6c3bd9d9`` |
| Signal-1M(RT) | [Homepage](https://research.signal-ai.com/datasets/signal1m-tweetir.html)| ``signal1m`` | ``test``| 97 | 2.86M | 19.6 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#4-signal-1m) |
| TREC-NEWS | [Homepage](https://trec.nist.gov/data/news2019.html) | ``trec-news`` | ``test``| 57 | 595K | 19.6 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#1-trec-news) |
| ArguAna | [Homepage](http://argumentation.bplaced.net/arguana/data) | ``arguana``| ``test`` | 1,406 | 8.67K | 1.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/arguana.zip) | ``8ad3e3c2a5867cdced806d6503f29b99`` |
| Touche-2020| [Homepage](https://webis.de/events/touche-20/shared-task-1.html) | ``webis-touche2020``| ``test``| 49 | 382K | 19.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/webis-touche2020.zip) | ``46f650ba5a527fc69e0a6521c5a23563`` |
| CQADupstack| [Homepage](http://nlp.cis.unimelb.edu.au/resources/cqadupstack/) | ``cqadupstack``| ``test``| 13,145 | 457K | 1.4 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/cqadupstack.zip) | ``4e41456d7df8ee7760a7f866133bda78`` |
| Quora| [Homepage](https://www.quora.com/q/quoradata/First-Quora-Dataset-Release-Question-Pairs) | ``quora``| ``dev``<br>``test``| 10,000 | 523K | 1.6 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/quora.zip) | ``18fb154900ba42a600f84b839c173167`` |
| DBPedia | [Homepage](https://github.com/iai-group/DBpedia-Entity/) | ``dbpedia-entity``| ``dev``<br>``test``| 400 | 4.63M | 38.2 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/dbpedia-entity.zip) | ``c2a39eb420a3164af735795df012ac2c`` |
| SCIDOCS| [Homepage](https://allenai.org/data/scidocs) | ``scidocs``| ``test``| 1,000 | 25K | 4.9 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/scidocs.zip) | ``38121350fc3a4d2f48850f6aff52e4a9`` |
| FEVER | [Homepage](http://fever.ai) | ``fever``| ``train``<br>``dev``<br>``test``| 6,666 | 5.42M | 1.2| [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/fever.zip) | ``5a818580227bfb4b35bb6fa46d9b6c03`` |
| Climate-FEVER| [Homepage](http://climatefever.ai) | ``climate-fever``|``test``| 1,535 | 5.42M | 3.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/climate-fever.zip) | ``8b66f0a9126c521bae2bde127b4dc99d`` |
| SciFact| [Homepage](https://github.com/allenai/scifact) | ``scifact``| ``train``<br>``test``| 300 | 5K | 1.1 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/scifact.zip) | ``5f7d1de60b170fc8027bb7898e2efca1`` |
| Robust04 | [Homepage](https://trec.nist.gov/data/robust/04.guidelines.html) | ``robust04``| ``test``| 249 | 528K | 69.9 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#3-robust04) |
## Dataset Creation
### Curation Rationale
[Needs More Information]
### Source Data
#### Initial Data Collection and Normalization
[Needs More Information]
#### Who are the source language producers?
[Needs More Information]
### Annotations
#### Annotation process
[Needs More Information]
#### Who are the annotators?
[Needs More Information]
### Personal and Sensitive Information
[Needs More Information]
## Considerations for Using the Data
### Social Impact of Dataset
[Needs More Information]
### Discussion of Biases
[Needs More Information]
### Other Known Limitations
[Needs More Information]
## Additional Information
### Dataset Curators
[Needs More Information]
### Licensing Information
[Needs More Information]
### Citation Information
Cite as:
```
@inproceedings{
thakur2021beir,
title={{BEIR}: A Heterogeneous Benchmark for Zero-shot Evaluation of Information Retrieval Models},
author={Nandan Thakur and Nils Reimers and Andreas R{\"u}ckl{\'e} and Abhishek Srivastava and Iryna Gurevych},
booktitle={Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2)},
year={2021},
url={https://openreview.net/forum?id=wCu6T5xFjeJ}
}
```
### Contributions
Thanks to [@Nthakur20](https://github.com/Nthakur20) for adding this dataset. |
false |
# Dataset Card for BEIR Benchmark
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Homepage:** https://github.com/UKPLab/beir
- **Repository:** https://github.com/UKPLab/beir
- **Paper:** https://openreview.net/forum?id=wCu6T5xFjeJ
- **Leaderboard:** https://docs.google.com/spreadsheets/d/1L8aACyPaXrL8iEelJLGqlMqXKPX2oSP_R10pZoy77Ns
- **Point of Contact:** nandan.thakur@uwaterloo.ca
### Dataset Summary
BEIR is a heterogeneous benchmark that has been built from 18 diverse datasets representing 9 information retrieval tasks:
- Fact-checking: [FEVER](http://fever.ai), [Climate-FEVER](http://climatefever.ai), [SciFact](https://github.com/allenai/scifact)
- Question-Answering: [NQ](https://ai.google.com/research/NaturalQuestions), [HotpotQA](https://hotpotqa.github.io), [FiQA-2018](https://sites.google.com/view/fiqa/)
- Bio-Medical IR: [TREC-COVID](https://ir.nist.gov/covidSubmit/index.html), [BioASQ](http://bioasq.org), [NFCorpus](https://www.cl.uni-heidelberg.de/statnlpgroup/nfcorpus/)
- News Retrieval: [TREC-NEWS](https://trec.nist.gov/data/news2019.html), [Robust04](https://trec.nist.gov/data/robust/04.guidelines.html)
- Argument Retrieval: [Touche-2020](https://webis.de/events/touche-20/shared-task-1.html), [ArguAna](tp://argumentation.bplaced.net/arguana/data)
- Duplicate Question Retrieval: [Quora](https://www.quora.com/q/quoradata/First-Quora-Dataset-Release-Question-Pairs), [CqaDupstack](http://nlp.cis.unimelb.edu.au/resources/cqadupstack/)
- Citation-Prediction: [SCIDOCS](https://allenai.org/data/scidocs)
- Tweet Retrieval: [Signal-1M](https://research.signal-ai.com/datasets/signal1m-tweetir.html)
- Entity Retrieval: [DBPedia](https://github.com/iai-group/DBpedia-Entity/)
All these datasets have been preprocessed and can be used for your experiments.
```python
```
### Supported Tasks and Leaderboards
The dataset supports a leaderboard that evaluates models against task-specific metrics such as F1 or EM, as well as their ability to retrieve supporting information from Wikipedia.
The current best performing models can be found [here](https://eval.ai/web/challenges/challenge-page/689/leaderboard/).
### Languages
All tasks are in English (`en`).
## Dataset Structure
All BEIR datasets must contain a corpus, queries and qrels (relevance judgments file). They must be in the following format:
- `corpus` file: a `.jsonl` file (jsonlines) that contains a list of dictionaries, each with three fields `_id` with unique document identifier, `title` with document title (optional) and `text` with document paragraph or passage. For example: `{"_id": "doc1", "title": "Albert Einstein", "text": "Albert Einstein was a German-born...."}`
- `queries` file: a `.jsonl` file (jsonlines) that contains a list of dictionaries, each with two fields `_id` with unique query identifier and `text` with query text. For example: `{"_id": "q1", "text": "Who developed the mass-energy equivalence formula?"}`
- `qrels` file: a `.tsv` file (tab-seperated) that contains three columns, i.e. the `query-id`, `corpus-id` and `score` in this order. Keep 1st row as header. For example: `q1 doc1 1`
### Data Instances
A high level example of any beir dataset:
```python
corpus = {
"doc1" : {
"title": "Albert Einstein",
"text": "Albert Einstein was a German-born theoretical physicist. who developed the theory of relativity, \
one of the two pillars of modern physics (alongside quantum mechanics). His work is also known for \
its influence on the philosophy of science. He is best known to the general public for his mass–energy \
equivalence formula E = mc2, which has been dubbed 'the world's most famous equation'. He received the 1921 \
Nobel Prize in Physics 'for his services to theoretical physics, and especially for his discovery of the law \
of the photoelectric effect', a pivotal step in the development of quantum theory."
},
"doc2" : {
"title": "", # Keep title an empty string if not present
"text": "Wheat beer is a top-fermented beer which is brewed with a large proportion of wheat relative to the amount of \
malted barley. The two main varieties are German Weißbier and Belgian witbier; other types include Lambic (made\
with wild yeast), Berliner Weisse (a cloudy, sour beer), and Gose (a sour, salty beer)."
},
}
queries = {
"q1" : "Who developed the mass-energy equivalence formula?",
"q2" : "Which beer is brewed with a large proportion of wheat?"
}
qrels = {
"q1" : {"doc1": 1},
"q2" : {"doc2": 1},
}
```
### Data Fields
Examples from all configurations have the following features:
### Corpus
- `corpus`: a `dict` feature representing the document title and passage text, made up of:
- `_id`: a `string` feature representing the unique document id
- `title`: a `string` feature, denoting the title of the document.
- `text`: a `string` feature, denoting the text of the document.
### Queries
- `queries`: a `dict` feature representing the query, made up of:
- `_id`: a `string` feature representing the unique query id
- `text`: a `string` feature, denoting the text of the query.
### Qrels
- `qrels`: a `dict` feature representing the query document relevance judgements, made up of:
- `_id`: a `string` feature representing the query id
- `_id`: a `string` feature, denoting the document id.
- `score`: a `int32` feature, denoting the relevance judgement between query and document.
### Data Splits
| Dataset | Website| BEIR-Name | Type | Queries | Corpus | Rel D/Q | Down-load | md5 |
| -------- | -----| ---------| --------- | ----------- | ---------| ---------| :----------: | :------:|
| MSMARCO | [Homepage](https://microsoft.github.io/msmarco/)| ``msmarco`` | ``train``<br>``dev``<br>``test``| 6,980 | 8.84M | 1.1 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/msmarco.zip) | ``444067daf65d982533ea17ebd59501e4`` |
| TREC-COVID | [Homepage](https://ir.nist.gov/covidSubmit/index.html)| ``trec-covid``| ``test``| 50| 171K| 493.5 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/trec-covid.zip) | ``ce62140cb23feb9becf6270d0d1fe6d1`` |
| NFCorpus | [Homepage](https://www.cl.uni-heidelberg.de/statnlpgroup/nfcorpus/) | ``nfcorpus`` | ``train``<br>``dev``<br>``test``| 323 | 3.6K | 38.2 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/nfcorpus.zip) | ``a89dba18a62ef92f7d323ec890a0d38d`` |
| BioASQ | [Homepage](http://bioasq.org) | ``bioasq``| ``train``<br>``test`` | 500 | 14.91M | 8.05 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#2-bioasq) |
| NQ | [Homepage](https://ai.google.com/research/NaturalQuestions) | ``nq``| ``train``<br>``test``| 3,452 | 2.68M | 1.2 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/nq.zip) | ``d4d3d2e48787a744b6f6e691ff534307`` |
| HotpotQA | [Homepage](https://hotpotqa.github.io) | ``hotpotqa``| ``train``<br>``dev``<br>``test``| 7,405 | 5.23M | 2.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/hotpotqa.zip) | ``f412724f78b0d91183a0e86805e16114`` |
| FiQA-2018 | [Homepage](https://sites.google.com/view/fiqa/) | ``fiqa`` | ``train``<br>``dev``<br>``test``| 648 | 57K | 2.6 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/fiqa.zip) | ``17918ed23cd04fb15047f73e6c3bd9d9`` |
| Signal-1M(RT) | [Homepage](https://research.signal-ai.com/datasets/signal1m-tweetir.html)| ``signal1m`` | ``test``| 97 | 2.86M | 19.6 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#4-signal-1m) |
| TREC-NEWS | [Homepage](https://trec.nist.gov/data/news2019.html) | ``trec-news`` | ``test``| 57 | 595K | 19.6 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#1-trec-news) |
| ArguAna | [Homepage](http://argumentation.bplaced.net/arguana/data) | ``arguana``| ``test`` | 1,406 | 8.67K | 1.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/arguana.zip) | ``8ad3e3c2a5867cdced806d6503f29b99`` |
| Touche-2020| [Homepage](https://webis.de/events/touche-20/shared-task-1.html) | ``webis-touche2020``| ``test``| 49 | 382K | 19.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/webis-touche2020.zip) | ``46f650ba5a527fc69e0a6521c5a23563`` |
| CQADupstack| [Homepage](http://nlp.cis.unimelb.edu.au/resources/cqadupstack/) | ``cqadupstack``| ``test``| 13,145 | 457K | 1.4 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/cqadupstack.zip) | ``4e41456d7df8ee7760a7f866133bda78`` |
| Quora| [Homepage](https://www.quora.com/q/quoradata/First-Quora-Dataset-Release-Question-Pairs) | ``quora``| ``dev``<br>``test``| 10,000 | 523K | 1.6 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/quora.zip) | ``18fb154900ba42a600f84b839c173167`` |
| DBPedia | [Homepage](https://github.com/iai-group/DBpedia-Entity/) | ``dbpedia-entity``| ``dev``<br>``test``| 400 | 4.63M | 38.2 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/dbpedia-entity.zip) | ``c2a39eb420a3164af735795df012ac2c`` |
| SCIDOCS| [Homepage](https://allenai.org/data/scidocs) | ``scidocs``| ``test``| 1,000 | 25K | 4.9 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/scidocs.zip) | ``38121350fc3a4d2f48850f6aff52e4a9`` |
| FEVER | [Homepage](http://fever.ai) | ``fever``| ``train``<br>``dev``<br>``test``| 6,666 | 5.42M | 1.2| [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/fever.zip) | ``5a818580227bfb4b35bb6fa46d9b6c03`` |
| Climate-FEVER| [Homepage](http://climatefever.ai) | ``climate-fever``|``test``| 1,535 | 5.42M | 3.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/climate-fever.zip) | ``8b66f0a9126c521bae2bde127b4dc99d`` |
| SciFact| [Homepage](https://github.com/allenai/scifact) | ``scifact``| ``train``<br>``test``| 300 | 5K | 1.1 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/scifact.zip) | ``5f7d1de60b170fc8027bb7898e2efca1`` |
| Robust04 | [Homepage](https://trec.nist.gov/data/robust/04.guidelines.html) | ``robust04``| ``test``| 249 | 528K | 69.9 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#3-robust04) |
## Dataset Creation
### Curation Rationale
[Needs More Information]
### Source Data
#### Initial Data Collection and Normalization
[Needs More Information]
#### Who are the source language producers?
[Needs More Information]
### Annotations
#### Annotation process
[Needs More Information]
#### Who are the annotators?
[Needs More Information]
### Personal and Sensitive Information
[Needs More Information]
## Considerations for Using the Data
### Social Impact of Dataset
[Needs More Information]
### Discussion of Biases
[Needs More Information]
### Other Known Limitations
[Needs More Information]
## Additional Information
### Dataset Curators
[Needs More Information]
### Licensing Information
[Needs More Information]
### Citation Information
Cite as:
```
@inproceedings{
thakur2021beir,
title={{BEIR}: A Heterogeneous Benchmark for Zero-shot Evaluation of Information Retrieval Models},
author={Nandan Thakur and Nils Reimers and Andreas R{\"u}ckl{\'e} and Abhishek Srivastava and Iryna Gurevych},
booktitle={Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2)},
year={2021},
url={https://openreview.net/forum?id=wCu6T5xFjeJ}
}
```
### Contributions
Thanks to [@Nthakur20](https://github.com/Nthakur20) for adding this dataset. |
false |
# Dataset Card for SRSD-Feynman (Medium set)
## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Homepage:**
- **Repository:** https://github.com/omron-sinicx/srsd-benchmark
- **Paper:** [Rethinking Symbolic Regression Datasets and Benchmarks for Scientific Discovery](https://arxiv.org/abs/2206.10540)
- **Point of Contact:** [Yoshitaka Ushiku](mailto:yoshitaka.ushiku@sinicx.com)
### Dataset Summary
Our SRSD (Feynman) datasets are designed to discuss the performance of Symbolic Regression for Scientific Discovery.
We carefully reviewed the properties of each formula and its variables in [the Feynman Symbolic Regression Database](https://space.mit.edu/home/tegmark/aifeynman.html) to design reasonably realistic sampling range of values so that our SRSD datasets can be used for evaluating the potential of SRSD such as whether or not an SR method con (re)discover physical laws from such datasets.
This is the ***Medium set*** of our SRSD-Feynman datasets, which consists of the following 40 different physics formulas:
[](https://huggingface.co/datasets/yoshitomo-matsubara/srsd-feynman_medium/resolve/main/problem_table.pdf)
More details of these datasets are provided in [the paper and its supplementary material](https://arxiv.org/abs/2206.10540).
### Supported Tasks and Leaderboards
Symbolic Regression
## Dataset Structure
### Data Instances
Tabular data + Ground-truth equation per equation
Tabular data: (num_samples, num_variables+1), where the last (rightmost) column indicate output of the target function for given variables.
Note that the number of variables (`num_variables`) varies from equation to equation.
Ground-truth equation: *pickled* symbolic representation (equation with symbols in sympy) of the target function.
### Data Fields
For each dataset, we have
1. train split (txt file, whitespace as a delimiter)
2. val split (txt file, whitespace as a delimiter)
3. test split (txt file, whitespace as a delimiter)
4. true equation (pickle file for sympy object)
### Data Splits
- train: 8,000 samples per equation
- val: 1,000 samples per equation
- test: 1,000 samples per equation
## Dataset Creation
### Curation Rationale
We chose target equations based on [the Feynman Symbolic Regression Database](https://space.mit.edu/home/tegmark/aifeynman.html).
### Annotations
#### Annotation process
We significantly revised the sampling range for each variable from the annotations in the Feynman Symbolic Regression Database.
First, we checked the properties of each variable and treat physical constants (e.g., light speed, gravitational constant) as constants.
Next, variable ranges were defined to correspond to each typical physics experiment to confirm the physical phenomenon for each equation.
In cases where a specific experiment is difficult to be assumed, ranges were set within which the corresponding physical phenomenon can be seen.
Generally, the ranges are set to be sampled on log scales within their orders as 10^2 in order to take both large and small changes in value as the order changes.
Variables such as angles, for which a linear distribution is expected are set to be sampled uniformly.
In addition, variables that take a specific sign were set to be sampled within that range.
#### Who are the annotators?
The main annotators are
- Naoya Chiba (@nchiba)
- Ryo Igarashi (@rigarash)
### Personal and Sensitive Information
N/A
## Considerations for Using the Data
### Social Impact of Dataset
We annotated this dataset, assuming typical physical experiments. The dataset will engage research on symbolic regression for scientific discovery (SRSD) and help researchers discuss the potential of symbolic regression methods towards data-driven scientific discovery.
### Discussion of Biases
Our choices of target equations are based on [the Feynman Symbolic Regression Database](https://space.mit.edu/home/tegmark/aifeynman.html), which are focused on a field of Physics.
### Other Known Limitations
Some variables used in our datasets indicate some numbers (counts), which should be treated as integer.
Due to the capacity of 32-bit integer, however, we treated some of such variables as float e.g., number of molecules (10^{23} - 10^{25})
## Additional Information
### Dataset Curators
The main curators are
- Naoya Chiba (@nchiba)
- Ryo Igarashi (@rigarash)
### Licensing Information
MIT License
### Citation Information
[[Preprint](https://arxiv.org/abs/2206.10540)]
```bibtex
@article{matsubara2022rethinking,
title={Rethinking Symbolic Regression Datasets and Benchmarks for Scientific Discovery},
author={Matsubara, Yoshitomo and Chiba, Naoya and Igarashi, Ryo and Ushiku, Yoshitaka},
journal={arXiv preprint arXiv:2206.10540},
year={2022}
}
```
### Contributions
Authors:
- Yoshitomo Matsubara (@yoshitomo-matsubara)
- Naoya Chiba (@nchiba)
- Ryo Igarashi (@rigarash)
- Yoshitaka Ushiku (@yushiku)
|
true | |
false |
# Dataset Card for MAWPS_ar
## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
### Dataset Summary
MAWPS: A Math Word Problem Repository
### Supported Tasks
Math Word Problem Solving
### Languages
Supports Arabic and English
## Dataset Structure
### Data Fields
- `text_en`: a `string` feature.
- `text_ar`: a `string` feature.
- `eqn`: a `string` feature.
### Data Splits
|train|validation|test|
|----:|---------:|---:|
| 3636| 1040| 520|
## Dataset Creation
### Curation Rationale
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Source Data
#### Initial Data Collection and Normalization
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
#### Who are the source language producers?
[Rik Koncel-Kedziorski**, Subhro Roy**, Aida Amini, Nate Kushman and Hannaneh Hajishirzi.](https://aclanthology.org/N16-1136.pdf)
### Annotations
#### Annotation process
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
#### Who are the annotators?
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Personal and Sensitive Information
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
## Considerations for Using the Data
### Social Impact of Dataset
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Discussion of Biases
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Other Known Limitations
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
## Additional Information
### Dataset Curators
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Licensing Information
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Citation Information
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Contributions
Special thanks to Associate Professor Marwan Torki and all my colleagues in CC491N (NLP) class for helping me translate this dataset. |
false |
# Dataset Card for BEIR Benchmark
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Homepage:** https://github.com/UKPLab/beir
- **Repository:** https://github.com/UKPLab/beir
- **Paper:** https://openreview.net/forum?id=wCu6T5xFjeJ
- **Leaderboard:** https://docs.google.com/spreadsheets/d/1L8aACyPaXrL8iEelJLGqlMqXKPX2oSP_R10pZoy77Ns
- **Point of Contact:** nandan.thakur@uwaterloo.ca
### Dataset Summary
BEIR is a heterogeneous benchmark that has been built from 18 diverse datasets representing 9 information retrieval tasks:
- Fact-checking: [FEVER](http://fever.ai), [Climate-FEVER](http://climatefever.ai), [SciFact](https://github.com/allenai/scifact)
- Question-Answering: [NQ](https://ai.google.com/research/NaturalQuestions), [HotpotQA](https://hotpotqa.github.io), [FiQA-2018](https://sites.google.com/view/fiqa/)
- Bio-Medical IR: [TREC-COVID](https://ir.nist.gov/covidSubmit/index.html), [BioASQ](http://bioasq.org), [NFCorpus](https://www.cl.uni-heidelberg.de/statnlpgroup/nfcorpus/)
- News Retrieval: [TREC-NEWS](https://trec.nist.gov/data/news2019.html), [Robust04](https://trec.nist.gov/data/robust/04.guidelines.html)
- Argument Retrieval: [Touche-2020](https://webis.de/events/touche-20/shared-task-1.html), [ArguAna](tp://argumentation.bplaced.net/arguana/data)
- Duplicate Question Retrieval: [Quora](https://www.quora.com/q/quoradata/First-Quora-Dataset-Release-Question-Pairs), [CqaDupstack](http://nlp.cis.unimelb.edu.au/resources/cqadupstack/)
- Citation-Prediction: [SCIDOCS](https://allenai.org/data/scidocs)
- Tweet Retrieval: [Signal-1M](https://research.signal-ai.com/datasets/signal1m-tweetir.html)
- Entity Retrieval: [DBPedia](https://github.com/iai-group/DBpedia-Entity/)
All these datasets have been preprocessed and can be used for your experiments.
```python
```
### Supported Tasks and Leaderboards
The dataset supports a leaderboard that evaluates models against task-specific metrics such as F1 or EM, as well as their ability to retrieve supporting information from Wikipedia.
The current best performing models can be found [here](https://eval.ai/web/challenges/challenge-page/689/leaderboard/).
### Languages
All tasks are in English (`en`).
## Dataset Structure
All BEIR datasets must contain a corpus, queries and qrels (relevance judgments file). They must be in the following format:
- `corpus` file: a `.jsonl` file (jsonlines) that contains a list of dictionaries, each with three fields `_id` with unique document identifier, `title` with document title (optional) and `text` with document paragraph or passage. For example: `{"_id": "doc1", "title": "Albert Einstein", "text": "Albert Einstein was a German-born...."}`
- `queries` file: a `.jsonl` file (jsonlines) that contains a list of dictionaries, each with two fields `_id` with unique query identifier and `text` with query text. For example: `{"_id": "q1", "text": "Who developed the mass-energy equivalence formula?"}`
- `qrels` file: a `.tsv` file (tab-seperated) that contains three columns, i.e. the `query-id`, `corpus-id` and `score` in this order. Keep 1st row as header. For example: `q1 doc1 1`
### Data Instances
A high level example of any beir dataset:
```python
corpus = {
"doc1" : {
"title": "Albert Einstein",
"text": "Albert Einstein was a German-born theoretical physicist. who developed the theory of relativity, \
one of the two pillars of modern physics (alongside quantum mechanics). His work is also known for \
its influence on the philosophy of science. He is best known to the general public for his mass–energy \
equivalence formula E = mc2, which has been dubbed 'the world's most famous equation'. He received the 1921 \
Nobel Prize in Physics 'for his services to theoretical physics, and especially for his discovery of the law \
of the photoelectric effect', a pivotal step in the development of quantum theory."
},
"doc2" : {
"title": "", # Keep title an empty string if not present
"text": "Wheat beer is a top-fermented beer which is brewed with a large proportion of wheat relative to the amount of \
malted barley. The two main varieties are German Weißbier and Belgian witbier; other types include Lambic (made\
with wild yeast), Berliner Weisse (a cloudy, sour beer), and Gose (a sour, salty beer)."
},
}
queries = {
"q1" : "Who developed the mass-energy equivalence formula?",
"q2" : "Which beer is brewed with a large proportion of wheat?"
}
qrels = {
"q1" : {"doc1": 1},
"q2" : {"doc2": 1},
}
```
### Data Fields
Examples from all configurations have the following features:
### Corpus
- `corpus`: a `dict` feature representing the document title and passage text, made up of:
- `_id`: a `string` feature representing the unique document id
- `title`: a `string` feature, denoting the title of the document.
- `text`: a `string` feature, denoting the text of the document.
### Queries
- `queries`: a `dict` feature representing the query, made up of:
- `_id`: a `string` feature representing the unique query id
- `text`: a `string` feature, denoting the text of the query.
### Qrels
- `qrels`: a `dict` feature representing the query document relevance judgements, made up of:
- `_id`: a `string` feature representing the query id
- `_id`: a `string` feature, denoting the document id.
- `score`: a `int32` feature, denoting the relevance judgement between query and document.
### Data Splits
| Dataset | Website| BEIR-Name | Type | Queries | Corpus | Rel D/Q | Down-load | md5 |
| -------- | -----| ---------| --------- | ----------- | ---------| ---------| :----------: | :------:|
| MSMARCO | [Homepage](https://microsoft.github.io/msmarco/)| ``msmarco`` | ``train``<br>``dev``<br>``test``| 6,980 | 8.84M | 1.1 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/msmarco.zip) | ``444067daf65d982533ea17ebd59501e4`` |
| TREC-COVID | [Homepage](https://ir.nist.gov/covidSubmit/index.html)| ``trec-covid``| ``test``| 50| 171K| 493.5 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/trec-covid.zip) | ``ce62140cb23feb9becf6270d0d1fe6d1`` |
| NFCorpus | [Homepage](https://www.cl.uni-heidelberg.de/statnlpgroup/nfcorpus/) | ``nfcorpus`` | ``train``<br>``dev``<br>``test``| 323 | 3.6K | 38.2 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/nfcorpus.zip) | ``a89dba18a62ef92f7d323ec890a0d38d`` |
| BioASQ | [Homepage](http://bioasq.org) | ``bioasq``| ``train``<br>``test`` | 500 | 14.91M | 8.05 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#2-bioasq) |
| NQ | [Homepage](https://ai.google.com/research/NaturalQuestions) | ``nq``| ``train``<br>``test``| 3,452 | 2.68M | 1.2 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/nq.zip) | ``d4d3d2e48787a744b6f6e691ff534307`` |
| HotpotQA | [Homepage](https://hotpotqa.github.io) | ``hotpotqa``| ``train``<br>``dev``<br>``test``| 7,405 | 5.23M | 2.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/hotpotqa.zip) | ``f412724f78b0d91183a0e86805e16114`` |
| FiQA-2018 | [Homepage](https://sites.google.com/view/fiqa/) | ``fiqa`` | ``train``<br>``dev``<br>``test``| 648 | 57K | 2.6 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/fiqa.zip) | ``17918ed23cd04fb15047f73e6c3bd9d9`` |
| Signal-1M(RT) | [Homepage](https://research.signal-ai.com/datasets/signal1m-tweetir.html)| ``signal1m`` | ``test``| 97 | 2.86M | 19.6 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#4-signal-1m) |
| TREC-NEWS | [Homepage](https://trec.nist.gov/data/news2019.html) | ``trec-news`` | ``test``| 57 | 595K | 19.6 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#1-trec-news) |
| ArguAna | [Homepage](http://argumentation.bplaced.net/arguana/data) | ``arguana``| ``test`` | 1,406 | 8.67K | 1.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/arguana.zip) | ``8ad3e3c2a5867cdced806d6503f29b99`` |
| Touche-2020| [Homepage](https://webis.de/events/touche-20/shared-task-1.html) | ``webis-touche2020``| ``test``| 49 | 382K | 19.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/webis-touche2020.zip) | ``46f650ba5a527fc69e0a6521c5a23563`` |
| CQADupstack| [Homepage](http://nlp.cis.unimelb.edu.au/resources/cqadupstack/) | ``cqadupstack``| ``test``| 13,145 | 457K | 1.4 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/cqadupstack.zip) | ``4e41456d7df8ee7760a7f866133bda78`` |
| Quora| [Homepage](https://www.quora.com/q/quoradata/First-Quora-Dataset-Release-Question-Pairs) | ``quora``| ``dev``<br>``test``| 10,000 | 523K | 1.6 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/quora.zip) | ``18fb154900ba42a600f84b839c173167`` |
| DBPedia | [Homepage](https://github.com/iai-group/DBpedia-Entity/) | ``dbpedia-entity``| ``dev``<br>``test``| 400 | 4.63M | 38.2 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/dbpedia-entity.zip) | ``c2a39eb420a3164af735795df012ac2c`` |
| SCIDOCS| [Homepage](https://allenai.org/data/scidocs) | ``scidocs``| ``test``| 1,000 | 25K | 4.9 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/scidocs.zip) | ``38121350fc3a4d2f48850f6aff52e4a9`` |
| FEVER | [Homepage](http://fever.ai) | ``fever``| ``train``<br>``dev``<br>``test``| 6,666 | 5.42M | 1.2| [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/fever.zip) | ``5a818580227bfb4b35bb6fa46d9b6c03`` |
| Climate-FEVER| [Homepage](http://climatefever.ai) | ``climate-fever``|``test``| 1,535 | 5.42M | 3.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/climate-fever.zip) | ``8b66f0a9126c521bae2bde127b4dc99d`` |
| SciFact| [Homepage](https://github.com/allenai/scifact) | ``scifact``| ``train``<br>``test``| 300 | 5K | 1.1 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/scifact.zip) | ``5f7d1de60b170fc8027bb7898e2efca1`` |
| Robust04 | [Homepage](https://trec.nist.gov/data/robust/04.guidelines.html) | ``robust04``| ``test``| 249 | 528K | 69.9 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#3-robust04) |
## Dataset Creation
### Curation Rationale
[Needs More Information]
### Source Data
#### Initial Data Collection and Normalization
[Needs More Information]
#### Who are the source language producers?
[Needs More Information]
### Annotations
#### Annotation process
[Needs More Information]
#### Who are the annotators?
[Needs More Information]
### Personal and Sensitive Information
[Needs More Information]
## Considerations for Using the Data
### Social Impact of Dataset
[Needs More Information]
### Discussion of Biases
[Needs More Information]
### Other Known Limitations
[Needs More Information]
## Additional Information
### Dataset Curators
[Needs More Information]
### Licensing Information
[Needs More Information]
### Citation Information
Cite as:
```
@inproceedings{
thakur2021beir,
title={{BEIR}: A Heterogeneous Benchmark for Zero-shot Evaluation of Information Retrieval Models},
author={Nandan Thakur and Nils Reimers and Andreas R{\"u}ckl{\'e} and Abhishek Srivastava and Iryna Gurevych},
booktitle={Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2)},
year={2021},
url={https://openreview.net/forum?id=wCu6T5xFjeJ}
}
```
### Contributions
Thanks to [@Nthakur20](https://github.com/Nthakur20) for adding this dataset. |
false |
# Dataset Card for BEIR Benchmark
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Homepage:** https://github.com/UKPLab/beir
- **Repository:** https://github.com/UKPLab/beir
- **Paper:** https://openreview.net/forum?id=wCu6T5xFjeJ
- **Leaderboard:** https://docs.google.com/spreadsheets/d/1L8aACyPaXrL8iEelJLGqlMqXKPX2oSP_R10pZoy77Ns
- **Point of Contact:** nandan.thakur@uwaterloo.ca
### Dataset Summary
BEIR is a heterogeneous benchmark that has been built from 18 diverse datasets representing 9 information retrieval tasks:
- Fact-checking: [FEVER](http://fever.ai), [Climate-FEVER](http://climatefever.ai), [SciFact](https://github.com/allenai/scifact)
- Question-Answering: [NQ](https://ai.google.com/research/NaturalQuestions), [HotpotQA](https://hotpotqa.github.io), [FiQA-2018](https://sites.google.com/view/fiqa/)
- Bio-Medical IR: [TREC-COVID](https://ir.nist.gov/covidSubmit/index.html), [BioASQ](http://bioasq.org), [NFCorpus](https://www.cl.uni-heidelberg.de/statnlpgroup/nfcorpus/)
- News Retrieval: [TREC-NEWS](https://trec.nist.gov/data/news2019.html), [Robust04](https://trec.nist.gov/data/robust/04.guidelines.html)
- Argument Retrieval: [Touche-2020](https://webis.de/events/touche-20/shared-task-1.html), [ArguAna](tp://argumentation.bplaced.net/arguana/data)
- Duplicate Question Retrieval: [Quora](https://www.quora.com/q/quoradata/First-Quora-Dataset-Release-Question-Pairs), [CqaDupstack](http://nlp.cis.unimelb.edu.au/resources/cqadupstack/)
- Citation-Prediction: [SCIDOCS](https://allenai.org/data/scidocs)
- Tweet Retrieval: [Signal-1M](https://research.signal-ai.com/datasets/signal1m-tweetir.html)
- Entity Retrieval: [DBPedia](https://github.com/iai-group/DBpedia-Entity/)
All these datasets have been preprocessed and can be used for your experiments.
```python
```
### Supported Tasks and Leaderboards
The dataset supports a leaderboard that evaluates models against task-specific metrics such as F1 or EM, as well as their ability to retrieve supporting information from Wikipedia.
The current best performing models can be found [here](https://eval.ai/web/challenges/challenge-page/689/leaderboard/).
### Languages
All tasks are in English (`en`).
## Dataset Structure
All BEIR datasets must contain a corpus, queries and qrels (relevance judgments file). They must be in the following format:
- `corpus` file: a `.jsonl` file (jsonlines) that contains a list of dictionaries, each with three fields `_id` with unique document identifier, `title` with document title (optional) and `text` with document paragraph or passage. For example: `{"_id": "doc1", "title": "Albert Einstein", "text": "Albert Einstein was a German-born...."}`
- `queries` file: a `.jsonl` file (jsonlines) that contains a list of dictionaries, each with two fields `_id` with unique query identifier and `text` with query text. For example: `{"_id": "q1", "text": "Who developed the mass-energy equivalence formula?"}`
- `qrels` file: a `.tsv` file (tab-seperated) that contains three columns, i.e. the `query-id`, `corpus-id` and `score` in this order. Keep 1st row as header. For example: `q1 doc1 1`
### Data Instances
A high level example of any beir dataset:
```python
corpus = {
"doc1" : {
"title": "Albert Einstein",
"text": "Albert Einstein was a German-born theoretical physicist. who developed the theory of relativity, \
one of the two pillars of modern physics (alongside quantum mechanics). His work is also known for \
its influence on the philosophy of science. He is best known to the general public for his mass–energy \
equivalence formula E = mc2, which has been dubbed 'the world's most famous equation'. He received the 1921 \
Nobel Prize in Physics 'for his services to theoretical physics, and especially for his discovery of the law \
of the photoelectric effect', a pivotal step in the development of quantum theory."
},
"doc2" : {
"title": "", # Keep title an empty string if not present
"text": "Wheat beer is a top-fermented beer which is brewed with a large proportion of wheat relative to the amount of \
malted barley. The two main varieties are German Weißbier and Belgian witbier; other types include Lambic (made\
with wild yeast), Berliner Weisse (a cloudy, sour beer), and Gose (a sour, salty beer)."
},
}
queries = {
"q1" : "Who developed the mass-energy equivalence formula?",
"q2" : "Which beer is brewed with a large proportion of wheat?"
}
qrels = {
"q1" : {"doc1": 1},
"q2" : {"doc2": 1},
}
```
### Data Fields
Examples from all configurations have the following features:
### Corpus
- `corpus`: a `dict` feature representing the document title and passage text, made up of:
- `_id`: a `string` feature representing the unique document id
- `title`: a `string` feature, denoting the title of the document.
- `text`: a `string` feature, denoting the text of the document.
### Queries
- `queries`: a `dict` feature representing the query, made up of:
- `_id`: a `string` feature representing the unique query id
- `text`: a `string` feature, denoting the text of the query.
### Qrels
- `qrels`: a `dict` feature representing the query document relevance judgements, made up of:
- `_id`: a `string` feature representing the query id
- `_id`: a `string` feature, denoting the document id.
- `score`: a `int32` feature, denoting the relevance judgement between query and document.
### Data Splits
| Dataset | Website| BEIR-Name | Type | Queries | Corpus | Rel D/Q | Down-load | md5 |
| -------- | -----| ---------| --------- | ----------- | ---------| ---------| :----------: | :------:|
| MSMARCO | [Homepage](https://microsoft.github.io/msmarco/)| ``msmarco`` | ``train``<br>``dev``<br>``test``| 6,980 | 8.84M | 1.1 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/msmarco.zip) | ``444067daf65d982533ea17ebd59501e4`` |
| TREC-COVID | [Homepage](https://ir.nist.gov/covidSubmit/index.html)| ``trec-covid``| ``test``| 50| 171K| 493.5 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/trec-covid.zip) | ``ce62140cb23feb9becf6270d0d1fe6d1`` |
| NFCorpus | [Homepage](https://www.cl.uni-heidelberg.de/statnlpgroup/nfcorpus/) | ``nfcorpus`` | ``train``<br>``dev``<br>``test``| 323 | 3.6K | 38.2 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/nfcorpus.zip) | ``a89dba18a62ef92f7d323ec890a0d38d`` |
| BioASQ | [Homepage](http://bioasq.org) | ``bioasq``| ``train``<br>``test`` | 500 | 14.91M | 8.05 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#2-bioasq) |
| NQ | [Homepage](https://ai.google.com/research/NaturalQuestions) | ``nq``| ``train``<br>``test``| 3,452 | 2.68M | 1.2 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/nq.zip) | ``d4d3d2e48787a744b6f6e691ff534307`` |
| HotpotQA | [Homepage](https://hotpotqa.github.io) | ``hotpotqa``| ``train``<br>``dev``<br>``test``| 7,405 | 5.23M | 2.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/hotpotqa.zip) | ``f412724f78b0d91183a0e86805e16114`` |
| FiQA-2018 | [Homepage](https://sites.google.com/view/fiqa/) | ``fiqa`` | ``train``<br>``dev``<br>``test``| 648 | 57K | 2.6 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/fiqa.zip) | ``17918ed23cd04fb15047f73e6c3bd9d9`` |
| Signal-1M(RT) | [Homepage](https://research.signal-ai.com/datasets/signal1m-tweetir.html)| ``signal1m`` | ``test``| 97 | 2.86M | 19.6 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#4-signal-1m) |
| TREC-NEWS | [Homepage](https://trec.nist.gov/data/news2019.html) | ``trec-news`` | ``test``| 57 | 595K | 19.6 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#1-trec-news) |
| ArguAna | [Homepage](http://argumentation.bplaced.net/arguana/data) | ``arguana``| ``test`` | 1,406 | 8.67K | 1.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/arguana.zip) | ``8ad3e3c2a5867cdced806d6503f29b99`` |
| Touche-2020| [Homepage](https://webis.de/events/touche-20/shared-task-1.html) | ``webis-touche2020``| ``test``| 49 | 382K | 19.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/webis-touche2020.zip) | ``46f650ba5a527fc69e0a6521c5a23563`` |
| CQADupstack| [Homepage](http://nlp.cis.unimelb.edu.au/resources/cqadupstack/) | ``cqadupstack``| ``test``| 13,145 | 457K | 1.4 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/cqadupstack.zip) | ``4e41456d7df8ee7760a7f866133bda78`` |
| Quora| [Homepage](https://www.quora.com/q/quoradata/First-Quora-Dataset-Release-Question-Pairs) | ``quora``| ``dev``<br>``test``| 10,000 | 523K | 1.6 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/quora.zip) | ``18fb154900ba42a600f84b839c173167`` |
| DBPedia | [Homepage](https://github.com/iai-group/DBpedia-Entity/) | ``dbpedia-entity``| ``dev``<br>``test``| 400 | 4.63M | 38.2 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/dbpedia-entity.zip) | ``c2a39eb420a3164af735795df012ac2c`` |
| SCIDOCS| [Homepage](https://allenai.org/data/scidocs) | ``scidocs``| ``test``| 1,000 | 25K | 4.9 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/scidocs.zip) | ``38121350fc3a4d2f48850f6aff52e4a9`` |
| FEVER | [Homepage](http://fever.ai) | ``fever``| ``train``<br>``dev``<br>``test``| 6,666 | 5.42M | 1.2| [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/fever.zip) | ``5a818580227bfb4b35bb6fa46d9b6c03`` |
| Climate-FEVER| [Homepage](http://climatefever.ai) | ``climate-fever``|``test``| 1,535 | 5.42M | 3.0 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/climate-fever.zip) | ``8b66f0a9126c521bae2bde127b4dc99d`` |
| SciFact| [Homepage](https://github.com/allenai/scifact) | ``scifact``| ``train``<br>``test``| 300 | 5K | 1.1 | [Link](https://public.ukp.informatik.tu-darmstadt.de/thakur/BEIR/datasets/scifact.zip) | ``5f7d1de60b170fc8027bb7898e2efca1`` |
| Robust04 | [Homepage](https://trec.nist.gov/data/robust/04.guidelines.html) | ``robust04``| ``test``| 249 | 528K | 69.9 | No | [How to Reproduce?](https://github.com/UKPLab/beir/blob/main/examples/dataset#3-robust04) |
## Dataset Creation
### Curation Rationale
[Needs More Information]
### Source Data
#### Initial Data Collection and Normalization
[Needs More Information]
#### Who are the source language producers?
[Needs More Information]
### Annotations
#### Annotation process
[Needs More Information]
#### Who are the annotators?
[Needs More Information]
### Personal and Sensitive Information
[Needs More Information]
## Considerations for Using the Data
### Social Impact of Dataset
[Needs More Information]
### Discussion of Biases
[Needs More Information]
### Other Known Limitations
[Needs More Information]
## Additional Information
### Dataset Curators
[Needs More Information]
### Licensing Information
[Needs More Information]
### Citation Information
Cite as:
```
@inproceedings{
thakur2021beir,
title={{BEIR}: A Heterogeneous Benchmark for Zero-shot Evaluation of Information Retrieval Models},
author={Nandan Thakur and Nils Reimers and Andreas R{\"u}ckl{\'e} and Abhishek Srivastava and Iryna Gurevych},
booktitle={Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2)},
year={2021},
url={https://openreview.net/forum?id=wCu6T5xFjeJ}
}
```
### Contributions
Thanks to [@Nthakur20](https://github.com/Nthakur20) for adding this dataset. |
false |
# Dataset Card for CLIP-Kinetics70
## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Dataset Preprocessing](#dataset-preprocessing)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Source Data](#source-data)
- [Simple Experiments](#dataset-creation)
- [Zero-shot Evaluation](#zero-shot)
- [Linear-probe Evaluation](#zero-shot)
## Dataset Description
### Dataset Summary
CLIP-Kinetics700 is a compressed version of the Kinetics700 dataset using OpenAI's CLIP model.
The original dataset is ~700 GB making it difficult to use and hold in memory on one machine. By downsampling each video to 1 FPS and encoding the frames using CLIP we we're able to compress the dataset to ~8 GB making it very memory-friendly and easy to use.
### Dataset Preprocessing
[clip-video-encode](https://github.com/iejMac/clip-video-encode) is a tool you can use to easily and efficiently compute CLIP embeddings from video frames. We used it to generate the embeddings for this dataset.
## Dataset Structure
### Data Format
We formatted this as a [WebDataset](https://github.com/webdataset/webdataset) for better data-loading performance when training the models.
Each split contains a list of tar files each with 10000 data samples. This format can be read and used easily using the EmbeddingWebDatasetReader from [clip-video-encode](https://github.com/iejMac/clip-video-encode).
```
CLIP-Kinetics700
├── splits.csv
├── ds_00000.tar
| ├── vid_00000.npy
| ├── vid_00000.txt
| ├── vid_00000.json
| ├── vid_00001.npy
| ├── vid_00001.txt
| ├── vid_00001.json
| └── ...
| ├── vid_10000.npy
| ├── vid_10000.txt
| ├── vid_10000.json
├── ds_00001.tar
| ├── vid_10001.npy
| ├── vid_10001.txt
| ├── vid_10001.json
│ ...
...
```
### Data Fields
* vid.npy: the numpy array with the per-frame embeddings. Shape -> (n_frames, 512)
* vid.cap: the "caption" of the video. In this case it is the Kinetics700 label.
* vid.json: additional metadata - YouTube video ID, start time, end time.
### Data Splits
* Train - 536489 samples | 54 tar's
* Validation - 33966 samples | 4 tar's
* Test - 64532 samples | 7 tar's
## Dataset Creation
### Source Data
Data was sourced from DeepMind's [Kinetics700](https://www.deepmind.com/open-source/kinetics) dataset and downloaded using [this](https://github.com/cvdfoundation/kinetics-dataset) convenient repository.
## Simple Experiments
Using [this repository](https://github.com/LAION-AI/temporal-embedding-aggregation) we evaluate CLIP-Kinetics700 with the following simple methods:
### [Zero-shot Evaluation](https://github.com/LAION-AI/temporal-embedding-aggregation/blob/master/src/evaluation/zero_shot.py)
| | Accuracy |
| ---------------- | -------- |
| Top-1 | 0.31 |
| Top-5 | 0.56 |
| mean(Top1, Top5) | 0.44 |
### [Linear-probe Evaluation](https://github.com/LAION-AI/temporal-embedding-aggregation/blob/master/src/evaluation/linear_probe.py)
| | Accuracy |
| ---------------- | -------- |
| Top-1 | 0.41 |
| Top-5 | 0.65 |
| mean(Top1, Top5) | 0.53 |
|
false |
# Dataset Card for Object Detection for Chess Pieces
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-instances)
- [Data Splits](#data-instances)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
## Dataset Description
- **Homepage:** https://github.com/faizankshaikh/chessDetection
- **Repository:** https://github.com/faizankshaikh/chessDetection
- **Paper:** -
- **Leaderboard:** -
- **Point of Contact:** [Faizan Shaikh](mailto:faizankshaikh@gmail.com)
### Dataset Summary
The "Object Detection for Chess Pieces" dataset is a toy dataset created (as suggested by the name!) to introduce object detection in a beginner friendly way. It is structured in a one object-one image manner, with the objects being of four classes, namely, Black King, White King, Black Queen and White Queen
### Supported Tasks and Leaderboards
- `object-detection`: The dataset can be used to train and evaluate simplistic object detection models
### Languages
The text (labels) in the dataset is in English
## Dataset Structure
### Data Instances
A data point comprises an image and the corresponding objects in bounding boxes.
```
{
'image': <PIL.PngImagePlugin.PngImageFile image mode=RGB size=224x224 at 0x23557C66160>,
'objects': { "label": [ 0 ], "bbox": [ [ 151, 151, 26, 26 ] ] }
}
```
### Data Fields
- `image`: A `PIL.Image.Image` object containing the 224x224 image.
- `label`: An integer between 0 and 3 representing the classes with the following mapping:
| Label | Description |
| --- | --- |
| 0 | blackKing |
| 1 | blackQueen |
| 2 | whiteKing |
| 3 | whiteQueen |
- `bbox`: A list of integers having sequence [x_center, y_center, width, height] for a particular bounding box
### Data Splits
The data is split into training and validation set. The training set contains 204 images and the validation set 52 images.
## Dataset Creation
### Curation Rationale
The dataset was created to be a simple benchmark for object detection
### Source Data
#### Initial Data Collection and Normalization
The data is obtained by machine generating images from "python-chess" library. Please refer [this code](https://github.com/faizankshaikh/chessDetection/blob/main/code/1.1%20create_images_with_labels.ipynb) to understand data generation pipeline
#### Who are the source language producers?
[Needs More Information]
### Annotations
#### Annotation process
The annotations were done manually.
#### Who are the annotators?
The annotations were done manually.
### Personal and Sensitive Information
None
## Considerations for Using the Data
### Social Impact of Dataset
The dataset can be considered as a beginner-friendly toy dataset for object detection. It should not be used for benchmarking state of the art object detection models, or be used for a deployed model.
### Discussion of Biases
[Needs More Information]
### Other Known Limitations
The dataset only contains four classes for simplicity. The complexity can be increased by considering all types of chess pieces, and by making it a multi-object detection problem
## Additional Information
### Dataset Curators
The dataset was created by Faizan Shaikh
### Licensing Information
The dataset is licensed as CC-BY-SA:2.0
### Citation Information
[Needs More Information] |
false |
# LibriS2S
This repo contains scripts and alignment data to create a dataset build further upon [librivoxDeEn](https://www.cl.uni-heidelberg.de/statnlpgroup/librivoxdeen/) such that it contains (German audio, German transcription, English audio, English transcription) quadruplets and can be used for Speech-to-Speech translation research. Because of this, the alignments are released under the same [Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License](https://creativecommons.org/licenses/by-nc-sa/4.0/) <div>
These alignments were collected by downloading the English audiobooks and using [aeneas](https://github.com/readbeyond/aeneas) to align the book chapters to the transcripts. For more information read the original [paper](https://arxiv.org/abs/2204.10593) (Presented at LREC 2022)
### The data
The English/German audio are available in the folder EN/DE respectively and can be downloaded from [this onedrive](https://onedrive.live.com/embed?cid=DCE49ACC2BDA7D8C&resid=DCE49ACC2BDA7D8C%2115663&authkey=ANmUz8gRUoyxmjk). In case there are any problems with the download, feel free to open an issue here or on [GitHub](https://github.com/PedroDKE/LibriS2S). <br/>
The repo structure is as follow:
- Alignments : Contains all the alignments for each book and chapter
- DE : Contains the German audio for each chapter per book.
- EN : Contains the English audio for each chapter per book.
- Example : contains example files on for the scraping and aligning explanations that were used to build this dataset.
- LibrivoxDeEn_alignments : Contains the base alignments from the LibrivoxDeEn dataset. <br/>
In case you feel a part of the data is missing, feel free to open an issue!
The full zipfile is about 52 GB of size.
### Scraping a book from Librivox
To download all chapters from a librivox url the following command can be used:
```
python scrape_audio_from_librivox.py \
--url https://librivox.org/undine-by-friedrich-de-la-motte-fouque/ \
--save_dir ./examples
```
### Allign a book from Librivox with the text from LibrivoxDeEn
To allign the previously downloaded book with the txt files and tsv tables provided by LibrivoxDeEn the following command, based on the example provided with this repo, can be used:
```
python align_text_and_audio.py \
--text_dir ./example/en_text/ \
--audio_path ./example/audio_chapters/ \
--aeneas_path ./example/aeneas/ \
--en_audio_export_path ./example/sentence_level_audio/ \
--total_alignment_path ./example/bi-lingual-alignment/ \
--librivoxdeen_alignment ./example/undine_data.tsv \
--aeneas_head_max 120 \
--aeneas_tail_min 5 \
```
**note:** the example folder in this repo already contains the first two chapters from [Undine](https://librivox.org/undine-by-friedrich-de-la-motte-fouque/) scraped from librivox and their transcripts and (modified to only contain the first 2 chapters) tsv table retrieved from LibrivoxDeEn.
Additional data to align can be scraped by using the same file shown previously and combined with the provided data from LibriVoxDeEn
Additionally with this repo the full alignment for the 8 following books with following LibrivoxDeEn id's are also given:
[9](https://librivox.org/the-picture-of-dorian-gray-1891-version-by-oscar-wilde/), [10](https://librivox.org/pandoras-box-by-frank-wedekind/), [13](https://librivox.org/survivors-of-the-chancellor-by-jules-verne/), [18](https://librivox.org/undine-by-friedrich-de-la-motte-fouque/), [23](https://librivox.org/around-the-world-in-80-days-by-jules-verne/), [108](https://librivox.org/elective-affinities-by-johann-wolfgang-von-goethe/), [110](https://librivox.org/candide-by-voltaire-3/), [120](https://librivox.org/the-metamorphosis-by-franz-kafka/).
Other books such as [11](https://librivox.org/the-castle-of-otranto-by-horace-walpole/), [36](https://librivox.org/the-rider-on-the-white-horse-by-theodor-storm/), [67](https://librivox.org/frankenstein-or-the-modern-prometheus-1818-by-mary-wollstonecraft-shelley/) and [54](https://librivox.org/white-nights-other-stories-by-fyodor-dostoyevsky/) are also inside of the librivoxDeEn dataset but the chapters do not correspond in a 1:1 mannner(for example: the German version of book 67 has 27 chapters but the English version has 29 and thus need to be re-aligned before the allignment script in this repo will work). Therefore these alignments are given but might have be different if you scrape them yourselves as the re-alignments might be different for you.
### Metrics on the alignment given in this repo.
Using the alignments given in this repo some metrics were collected and quickly displayed here, for this table and the next figure the books which were manually alligned, although provided in the zip, were not accounted for, but the full table can be found in the original paper.
| | German | English |
| :---: | :-: | :-: |
|number of files | 18868 | 18868 |
|total time (hh:mm:ss) | 39:11:08 | 40:52:31 |
|Speakers | 41 |22 |
note: the speakers were counted for each book seperatly so some speakers might be counter more than once.
the number of hours for each book aligned in this repo:<br>
<img src="https://user-images.githubusercontent.com/43861296/122250648-1f5f7f80-ceca-11eb-84fd-344a2261bf47.png" width="500">
when using this work, please cite the original paper and the LibrivoxDeEn authors
```
@inproceedings{jeuris-niehues-2022-libris2s,
title = "{L}ibri{S}2{S}: A {G}erman-{E}nglish Speech-to-Speech Translation Corpus",
author = "Jeuris, Pedro and
Niehues, Jan",
booktitle = "Proceedings of the Thirteenth Language Resources and Evaluation Conference",
month = jun,
year = "2022",
address = "Marseille, France",
publisher = "European Language Resources Association",
url = "https://aclanthology.org/2022.lrec-1.98",
pages = "928--935",
abstract = "Recently, we have seen an increasing interest in the area of speech-to-text translation. This has led to astonishing improvements in this area. In contrast, the activities in the area of speech-to-speech translation is still limited, although it is essential to overcome the language barrier. We believe that one of the limiting factors is the availability of appropriate training data. We address this issue by creating LibriS2S, to our knowledge the first publicly available speech-to-speech training corpus between German and English. For this corpus, we used independently created audio for German and English leading to an unbiased pronunciation of the text in both languages. This allows the creation of a new text-to-speech and speech-to-speech translation model that directly learns to generate the speech signal based on the pronunciation of the source language. Using this created corpus, we propose Text-to-Speech models based on the example of the recently proposed FastSpeech 2 model that integrates source language information. We do this by adapting the model to take information such as the pitch, energy or transcript from the source speech as additional input.",
}
```
```
@article{beilharz19,
title = {LibriVoxDeEn: A Corpus for German-to-English Speech Translation and Speech Recognition},
author = {Beilharz, Benjamin and Sun, Xin and Karimova, Sariya and Riezler, Stefan},
journal = {Proceedings of the Language Resources and Evaluation Conference},
journal-abbrev = {LREC},
year = {2020},
city = {Marseille, France},
url = {https://arxiv.org/pdf/1910.07924.pdf}
}
```
|
false |
MorisienMT is a dataset for Mauritian Creole Machine Translation. This dataset
consists of training, development and test set splits for English--Creole as
well as French--Creole translation. The data comes from a variety of sources
and hence can be considered as belonging to the general domain.
The training set for English--Creole contains 21,810 lines.
Finally, we also provide a Creole monolingual corpus of 45,364 lines.
Note that a significant portion of the dataset is a dictionary of word
pairs/triplets, nevertheless it is a start.
Feel free to use the dataset for your research but don't forget to attribute
our upcoming paper which will be uploaded to arxiv shortly.
NOTE: MorisienMT was originally partly developed by Dr Aneerav Sukhoo from the
University of Mauritius in 2014 when he was a visiting researcher in IIT
Bombay. Dr Sukhoo and Raj Dabre worked on the MT experiments together, but
never publicly released the dataset back then. Furthermore, the dataset splits
and experiments were not done in a highly principled manner, which is required
in the present day. Therefore, we improve the quality of splits and officially
release the data for people to use.
To use this dataset, request access via the [Surface Catalog](https://catalog.surfacedata.org/).
|
false |
# Dataset Card for Taskmaster-1
- **Repository:** https://github.com/google-research-datasets/Taskmaster/tree/master/TM-1-2019
- **Paper:** https://arxiv.org/pdf/1909.05358.pdf
- **Leaderboard:** None
- **Who transforms the dataset:** Qi Zhu(zhuq96 at gmail dot com)
To use this dataset, you need to install [ConvLab-3](https://github.com/ConvLab/ConvLab-3) platform first. Then you can load the dataset via:
```
from convlab.util import load_dataset, load_ontology, load_database
dataset = load_dataset('tm1')
ontology = load_ontology('tm1')
database = load_database('tm1')
```
For more usage please refer to [here](https://github.com/ConvLab/ConvLab-3/tree/master/data/unified_datasets).
### Dataset Summary
The original dataset consists of 13,215 task-based dialogs, including 5,507 spoken and 7,708 written dialogs created with two distinct procedures. Each conversation falls into one of six domains: ordering pizza, creating auto repair appointments, setting up ride service, ordering movie tickets, ordering coffee drinks and making restaurant reservations.
- **How to get the transformed data from original data:**
- Download [master.zip](https://github.com/google-research-datasets/Taskmaster/archive/refs/heads/master.zip).
- Run `python preprocess.py` in the current directory.
- **Main changes of the transformation:**
- Remove dialogs that are empty or only contain one speaker.
- Split woz-dialogs into train/validation/test randomly (8:1:1). The split of self-dialogs is followed the original dataset.
- Merge continuous turns by the same speaker (ignore repeated turns).
- Annotate `dialogue acts` according to the original segment annotations. Add `intent` annotation (inform/accept/reject). The type of `dialogue act` is set to `non-categorical` if the original segment annotation includes a specified `slot`. Otherwise, the type is set to `binary` (and the `slot` and `value` are empty) since it means general reference to a transaction, e.g. "OK your pizza has been ordered". If there are multiple spans overlapping, we only keep the shortest one, since we found that this simple strategy can reduce the noise in annotation.
- Add `domain`, `intent`, and `slot` descriptions.
- Add `state` by accumulate `non-categorical dialogue acts` in the order that they appear, except those whose intents are **reject**.
- Keep the first annotation since each conversation was annotated by two workers.
- **Annotations:**
- dialogue acts, state.
### Supported Tasks and Leaderboards
NLU, DST, Policy, NLG
### Languages
English
### Data Splits
| split | dialogues | utterances | avg_utt | avg_tokens | avg_domains | cat slot match(state) | cat slot match(goal) | cat slot match(dialogue act) | non-cat slot span(dialogue act) |
|------------|-------------|--------------|-----------|--------------|---------------|-------------------------|------------------------|--------------------------------|-----------------------------------|
| train | 10535 | 223322 | 21.2 | 8.75 | 1 | - | - | - | 100 |
| validation | 1318 | 27903 | 21.17 | 8.75 | 1 | - | - | - | 100 |
| test | 1322 | 27660 | 20.92 | 8.87 | 1 | - | - | - | 100 |
| all | 13175 | 278885 | 21.17 | 8.76 | 1 | - | - | - | 100 |
6 domains: ['uber_lyft', 'movie_ticket', 'restaurant_reservation', 'coffee_ordering', 'pizza_ordering', 'auto_repair']
- **cat slot match**: how many values of categorical slots are in the possible values of ontology in percentage.
- **non-cat slot span**: how many values of non-categorical slots have span annotation in percentage.
### Citation
```
@inproceedings{byrne-etal-2019-taskmaster,
title = {Taskmaster-1:Toward a Realistic and Diverse Dialog Dataset},
author = {Bill Byrne and Karthik Krishnamoorthi and Chinnadhurai Sankar and Arvind Neelakantan and Daniel Duckworth and Semih Yavuz and Ben Goodrich and Amit Dubey and Kyu-Young Kim and Andy Cedilnik},
booktitle = {2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing},
address = {Hong Kong},
year = {2019}
}
```
### Licensing Information
[**CC BY 4.0**](https://creativecommons.org/licenses/by/4.0/) |
false |
# Dataset Card for WOZ 2.0
- **Repository:** https://github.com/nmrksic/neural-belief-tracker/tree/master/data/woz
- **Paper:** https://aclanthology.org/P17-1163.pdf
- **Leaderboard:** None
- **Who transforms the dataset:** Qi Zhu(zhuq96 at gmail dot com)
To use this dataset, you need to install [ConvLab-3](https://github.com/ConvLab/ConvLab-3) platform first. Then you can load the dataset via:
```
from convlab.util import load_dataset, load_ontology, load_database
dataset = load_dataset('woz')
ontology = load_ontology('woz')
database = load_database('woz')
```
For more usage please refer to [here](https://github.com/ConvLab/ConvLab-3/tree/master/data/unified_datasets).
### Dataset Summary
Describe the dataset.
- **How to get the transformed data from original data:**
- download `woz_[train|validate|test]_en.json` from https://github.com/nmrksic/neural-belief-tracker/tree/master/data/woz and save to `woz` dir in the current directory.
- Run `python preprocess.py` in the current directory.
- **Main changes of the transformation:**
- domain is set to **restaurant**.
- normalize the value of categorical slots in state and dialogue acts.
- `belief_states` in WOZ dataset contains `request` intents, which are ignored in processing.
- use simple string match to find value spans of non-categorical slots.
- **Annotations:**
- User dialogue acts, state
### Supported Tasks and Leaderboards
NLU, DST, E2E
### Languages
English
### Data Splits
| split | dialogues | utterances | avg_utt | avg_tokens | avg_domains | cat slot match(state) | cat slot match(goal) | cat slot match(dialogue act) | non-cat slot span(dialogue act) |
|------------|-------------|--------------|-----------|--------------|---------------|-------------------------|------------------------|--------------------------------|-----------------------------------|
| train | 600 | 4472 | 7.45 | 11.37 | 1 | 100 | - | 100 | 96.56 |
| validation | 200 | 1460 | 7.3 | 11.28 | 1 | 100 | - | 100 | 95.52 |
| test | 400 | 2892 | 7.23 | 11.49 | 1 | 100 | - | 100 | 94.83 |
| all | 1200 | 8824 | 7.35 | 11.39 | 1 | 100 | - | 100 | 95.83 |
1 domains: ['restaurant']
- **cat slot match**: how many values of categorical slots are in the possible values of ontology in percentage.
- **non-cat slot span**: how many values of non-categorical slots have span annotation in percentage.
### Citation
```
@inproceedings{mrksic-etal-2017-neural,
title = "Neural Belief Tracker: Data-Driven Dialogue State Tracking",
author = "Mrk{\v{s}}i{\'c}, Nikola and
{\'O} S{\'e}aghdha, Diarmuid and
Wen, Tsung-Hsien and
Thomson, Blaise and
Young, Steve",
booktitle = "Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2017",
address = "Vancouver, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/P17-1163",
doi = "10.18653/v1/P17-1163",
pages = "1777--1788",
}
```
### Licensing Information
Apache License, Version 2.0
|
false |
# Dataset Card for Camrest
- **Repository:** https://www.repository.cam.ac.uk/handle/1810/260970
- **Paper:** https://aclanthology.org/D16-1233/
- **Leaderboard:** None
- **Who transforms the dataset:** Qi Zhu(zhuq96 at gmail dot com)
To use this dataset, you need to install [ConvLab-3](https://github.com/ConvLab/ConvLab-3) platform first. Then you can load the dataset via:
```
from convlab.util import load_dataset, load_ontology, load_database
dataset = load_dataset('camrest')
ontology = load_ontology('camrest')
database = load_database('camrest')
```
For more usage please refer to [here](https://github.com/ConvLab/ConvLab-3/tree/master/data/unified_datasets).
### Dataset Summary
Cambridge restaurant dialogue domain dataset collected for developing neural network based dialogue systems. The two papers published based on this dataset are: 1. A Network-based End-to-End Trainable Task-oriented Dialogue System 2. Conditional Generation and Snapshot Learning in Neural Dialogue Systems. The dataset was collected based on the Wizard of Oz experiment on Amazon MTurk. Each dialogue contains a goal label and several exchanges between a customer and the system. Each user turn was labelled by a set of slot-value pairs representing a coarse representation of dialogue state (`slu` field). There are in total 676 dialogue, in which most of the dialogues are finished but some of dialogues were not.
- **How to get the transformed data from original data:**
- Run `python preprocess.py` in the current directory. Need `../../camrest/` as the original data.
- **Main changes of the transformation:**
- Add dialogue act annotation according to the state change. This step was done by ConvLab-2 and we use the processed dialog acts here.
- Rename `pricerange` to `price range`
- Add character level span annotation for non-categorical slots.
- **Annotations:**
- user goal, dialogue acts, state.
### Supported Tasks and Leaderboards
NLU, DST, Policy, NLG, E2E, User simulator
### Languages
English
### Data Splits
| split | dialogues | utterances | avg_utt | avg_tokens | avg_domains | cat slot match(state) | cat slot match(goal) | cat slot match(dialogue act) | non-cat slot span(dialogue act) |
| ---------- | --------- | ---------- | ------- | ---------- | ----------- | --------------------- | -------------------- | ---------------------------- | ------------------------------- |
| train | 406 | 3342 | 8.23 | 10.6 | 1 | 100 | 100 | 100 | 99.83 |
| validation | 135 | 1076 | 7.97 | 11.26 | 1 | 100 | 100 | 100 | 100 |
| test | 135 | 1070 | 7.93 | 11.01 | 1 | 100 | 100 | 100 | 100 |
| all | 676 | 5488 | 8.12 | 10.81 | 1 | 100 | 100 | 100 | 99.9 |
1 domains: ['restaurant']
- **cat slot match**: how many values of categorical slots are in the possible values of ontology in percentage.
- **non-cat slot span**: how many values of non-categorical slots have span annotation in percentage.
### Citation
```
@inproceedings{wen-etal-2016-conditional,
title = "Conditional Generation and Snapshot Learning in Neural Dialogue Systems",
author = "Wen, Tsung-Hsien and Ga{\v{s}}i{\'c}, Milica and Mrk{\v{s}}i{\'c}, Nikola and Rojas-Barahona, Lina M. and Su, Pei-Hao and Ultes, Stefan and Vandyke, David and Young, Steve",
booktitle = "Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2016",
address = "Austin, Texas",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/D16-1233",
doi = "10.18653/v1/D16-1233",
pages = "2153--2162",
}
```
### Licensing Information
[**CC BY 4.0**](https://creativecommons.org/licenses/by/4.0/) |
false |
# Dataset Card for Taskmaster-2
- **Repository:** https://github.com/google-research-datasets/Taskmaster/tree/master/TM-2-2020
- **Paper:** https://arxiv.org/pdf/1909.05358.pdf
- **Leaderboard:** None
- **Who transforms the dataset:** Qi Zhu(zhuq96 at gmail dot com)
To use this dataset, you need to install [ConvLab-3](https://github.com/ConvLab/ConvLab-3) platform first. Then you can load the dataset via:
```
from convlab.util import load_dataset, load_ontology, load_database
dataset = load_dataset('tm2')
ontology = load_ontology('tm2')
database = load_database('tm2')
```
For more usage please refer to [here](https://github.com/ConvLab/ConvLab-3/tree/master/data/unified_datasets).
### Dataset Summary
The Taskmaster-2 dataset consists of 17,289 dialogs in the seven domains. Unlike Taskmaster-1, which includes both written "self-dialogs" and spoken two-person dialogs, Taskmaster-2 consists entirely of spoken two-person dialogs. In addition, while Taskmaster-1 is almost exclusively task-based, Taskmaster-2 contains a good number of search- and recommendation-oriented dialogs, as seen for example in the restaurants, flights, hotels, and movies verticals. The music browsing and sports conversations are almost exclusively search- and recommendation-based. All dialogs in this release were created using a Wizard of Oz (WOz) methodology in which crowdsourced workers played the role of a 'user' and trained call center operators played the role of the 'assistant'. In this way, users were led to believe they were interacting with an automated system that “spoke” using text-to-speech (TTS) even though it was in fact a human behind the scenes. As a result, users could express themselves however they chose in the context of an automated interface.
- **How to get the transformed data from original data:**
- Download [master.zip](https://github.com/google-research-datasets/Taskmaster/archive/refs/heads/master.zip).
- Run `python preprocess.py` in the current directory.
- **Main changes of the transformation:**
- Remove dialogs that are empty or only contain one speaker.
- Split each domain dialogs into train/validation/test randomly (8:1:1).
- Merge continuous turns by the same speaker (ignore repeated turns).
- Annotate `dialogue acts` according to the original segment annotations. Add `intent` annotation (`==inform`). The type of `dialogue act` is set to `non-categorical` if the `slot` is not in `anno2slot` in `preprocess.py`). Otherwise, the type is set to `binary` (and the `value` is empty). If there are multiple spans overlapping, we only keep the shortest one, since we found that this simple strategy can reduce the noise in annotation.
- Add `domain`, `intent`, and `slot` descriptions.
- Add `state` by accumulate `non-categorical dialogue acts` in the order that they appear.
- Keep the first annotation since each conversation was annotated by two workers.
- **Annotations:**
- dialogue acts, state.
### Supported Tasks and Leaderboards
NLU, DST, Policy, NLG
### Languages
English
### Data Splits
| split | dialogues | utterances | avg_utt | avg_tokens | avg_domains | cat slot match(state) | cat slot match(goal) | cat slot match(dialogue act) | non-cat slot span(dialogue act) |
|------------|-------------|--------------|-----------|--------------|---------------|-------------------------|------------------------|--------------------------------|-----------------------------------|
| train | 13838 | 234321 | 16.93 | 9.1 | 1 | - | - | - | 100 |
| validation | 1731 | 29349 | 16.95 | 9.15 | 1 | - | - | - | 100 |
| test | 1734 | 29447 | 16.98 | 9.07 | 1 | - | - | - | 100 |
| all | 17303 | 293117 | 16.94 | 9.1 | 1 | - | - | - | 100 |
7 domains: ['flights', 'food-ordering', 'hotels', 'movies', 'music', 'restaurant-search', 'sports']
- **cat slot match**: how many values of categorical slots are in the possible values of ontology in percentage.
- **non-cat slot span**: how many values of non-categorical slots have span annotation in percentage.
### Citation
```
@inproceedings{byrne-etal-2019-taskmaster,
title = {Taskmaster-1:Toward a Realistic and Diverse Dialog Dataset},
author = {Bill Byrne and Karthik Krishnamoorthi and Chinnadhurai Sankar and Arvind Neelakantan and Daniel Duckworth and Semih Yavuz and Ben Goodrich and Amit Dubey and Kyu-Young Kim and Andy Cedilnik},
booktitle = {2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing},
address = {Hong Kong},
year = {2019}
}
```
### Licensing Information
[**CC BY 4.0**](https://creativecommons.org/licenses/by/4.0/) |
false |
# Dataset Card for Taskmaster-3
- **Repository:** https://github.com/google-research-datasets/Taskmaster/tree/master/TM-3-2020
- **Paper:** https://aclanthology.org/2021.acl-long.55.pdf
- **Leaderboard:** None
- **Who transforms the dataset:** Qi Zhu(zhuq96 at gmail dot com)
To use this dataset, you need to install [ConvLab-3](https://github.com/ConvLab/ConvLab-3) platform first. Then you can load the dataset via:
```
from convlab.util import load_dataset, load_ontology, load_database
dataset = load_dataset('tm3')
ontology = load_ontology('tm3')
database = load_database('tm3')
```
For more usage please refer to [here](https://github.com/ConvLab/ConvLab-3/tree/master/data/unified_datasets).
### Dataset Summary
The Taskmaster-3 (aka TicketTalk) dataset consists of 23,789 movie ticketing dialogs (located in Taskmaster/TM-3-2020/data/). By "movie ticketing" we mean conversations where the customer's goal is to purchase tickets after deciding on theater, time, movie name, number of tickets, and date, or opt out of the transaction.
This collection was created using the "self-dialog" method. This means a single, crowd-sourced worker is paid to create a conversation writing turns for both speakers, i.e. the customer and the ticketing agent. In order to gather a wide range of conversational scenarios and linguistic phenomena, workers were given both open-ended as well as highly structured conversational tasks. In all, we used over three dozen sets of instructions while building this corpus. The "instructions" field in data.json provides the exact scenario workers were given to complete each dialog. In this way, conversations involve a wide variety of paths, from those where the customer decides on a movie based on genre, their location, current releases, or from what they already have in mind. In addition, dialogs also include error handling with repect to repair (e.g. "No, I said Tom Cruise."), clarifications (e.g. "Sorry. Did you want the AMC 16 or Century City 16?") and other common conversational hiccups. In some cases instructions are completely open ended e.g. "Pretend you are taking your friend to a movie in Salem, Oregon. Create a conversation where you end up buying two tickets after finding out what is playing in at least two local theaters. Make sure the ticket purchase includes a confirmation of the deatils by the agent before the purchase, including date, time, movie, theater, and number of tickets." In other cases we restrict the conversational content and structure by offering a partially completed conversation that the workers must finalize or fill in based a certain parameters. These partially completed dialogs are labeled "Auto template" in the "scenario" field shown for each conversation in the data.json file. In some cases, we provided a small KB from which workers would choose movies, theaters, etc. but in most cases (pre-pandemic) workers were told to use the internet to get accurate current details for their dialogs. In any case, all relevant entities are annotated.
- **How to get the transformed data from original data:**
- Download [master.zip](https://github.com/google-research-datasets/Taskmaster/archive/refs/heads/master.zip).
- Run `python preprocess.py` in the current directory.
- **Main changes of the transformation:**
- Remove dialogs that are empty or only contain one speaker.
- Split each domain dialogs into train/validation/test randomly (8:1:1).
- Merge continuous turns by the same speaker (ignore repeated turns).
- Annotate `dialogue acts` according to the original segment annotations. Add `intent` annotation (`==inform`). The type of `dialogue act` is set to `non-categorical` if the `slot` is not `description.other` or `description.plot`. Otherwise, the type is set to `binary` (and the `value` is empty). If there are multiple spans overlapping, we only keep the shortest one, since we found that this simple strategy can reduce the noise in annotation.
- Add `domain` and `intent` descriptions.
- Rename `api` to `db_results`.
- Add `state` by accumulate `non-categorical dialogue acts` in the order that they appear.
- **Annotations:**
- dialogue acts, state, db_results.
### Supported Tasks and Leaderboards
NLU, DST, Policy, NLG, E2E
### Languages
English
### Data Splits
| split | dialogues | utterances | avg_utt | avg_tokens | avg_domains | cat slot match(state) | cat slot match(goal) | cat slot match(dialogue act) | non-cat slot span(dialogue act) |
|------------|-------------|--------------|-----------|--------------|---------------|-------------------------|------------------------|--------------------------------|-----------------------------------|
| train | 18997 | 380646 | 20.04 | 10.48 | 1 | - | - | - | 100 |
| validation | 2380 | 47531 | 19.97 | 10.38 | 1 | - | - | - | 100 |
| test | 2380 | 48849 | 20.52 | 10.12 | 1 | - | - | - | 100 |
| all | 23757 | 477026 | 20.08 | 10.43 | 1 | - | - | - | 100 |
1 domains: ['movie']
- **cat slot match**: how many values of categorical slots are in the possible values of ontology in percentage.
- **non-cat slot span**: how many values of non-categorical slots have span annotation in percentage.
### Citation
```
@inproceedings{byrne-etal-2021-tickettalk,
title = "{T}icket{T}alk: Toward human-level performance with end-to-end, transaction-based dialog systems",
author = "Byrne, Bill and
Krishnamoorthi, Karthik and
Ganesh, Saravanan and
Kale, Mihir",
booktitle = "Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)",
month = aug,
year = "2021",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.acl-long.55",
doi = "10.18653/v1/2021.acl-long.55",
pages = "671--680",
}
```
### Licensing Information
[**CC BY 4.0**](https://creativecommons.org/licenses/by/4.0/) |
false |
# Dataset Card for DailyDialog
- **Repository:** http://yanran.li/dailydialog
- **Paper:** https://arxiv.org/pdf/1710.03957.pdf
- **Leaderboard:** None
- **Who transforms the dataset:** Qi Zhu(zhuq96 at gmail dot com)
To use this dataset, you need to install [ConvLab-3](https://github.com/ConvLab/ConvLab-3) platform first. Then you can load the dataset via:
```
from convlab.util import load_dataset, load_ontology, load_database
dataset = load_dataset('dailydialog')
ontology = load_ontology('dailydialog')
database = load_database('dailydialog')
```
For more usage please refer to [here](https://github.com/ConvLab/ConvLab-3/tree/master/data/unified_datasets).
### Dataset Summary
DailyDialog is a high-quality multi-turn dialog dataset. It is intriguing in several aspects. The language is human-written and less noisy. The dialogues in the dataset reflect our daily communication way and cover various topics about our daily life. We also manually label the developed dataset with communication intention and emotion information.
- **How to get the transformed data from original data:**
- Download [ijcnlp_dailydialog.zip](http://yanran.li/files/ijcnlp_dailydialog.zip).
- Run `python preprocess.py` in the current directory.
- **Main changes of the transformation:**
- Use `topic` annotation as `domain`. If duplicated dialogs are annotated with different topics, use the most frequent one.
- Use `intent` annotation as `binary` dialogue act.
- Retain emotion annotation in the `emotion` field of each turn.
- Use nltk to remove space before punctuation: `utt = ' '.join([detokenizer.detokenize(word_tokenize(s)) for s in sent_tokenize(utt)])`.
- Replace `" ’ "` with `"'"`: `utt = utt.replace(' ’ ', "'")`.
- Add space after full-stop
- **Annotations:**
- intent, emotion
### Supported Tasks and Leaderboards
NLU, NLG
### Languages
English
### Data Splits
| split | dialogues | utterances | avg_utt | avg_tokens | avg_domains | cat slot match(state) | cat slot match(goal) | cat slot match(dialogue act) | non-cat slot span(dialogue act) |
|------------|-------------|--------------|-----------|--------------|---------------|-------------------------|------------------------|--------------------------------|-----------------------------------|
| train | 11118 | 87170 | 7.84 | 11.22 | 1 | - | - | - | - |
| validation | 1000 | 8069 | 8.07 | 11.16 | 1 | - | - | - | - |
| test | 1000 | 7740 | 7.74 | 11.36 | 1 | - | - | - | - |
| all | 13118 | 102979 | 7.85 | 11.22 | 1 | - | - | - | - |
10 domains: ['Ordinary Life', 'School Life', 'Culture & Education', 'Attitude & Emotion', 'Relationship', 'Tourism', 'Health', 'Work', 'Politics', 'Finance']
- **cat slot match**: how many values of categorical slots are in the possible values of ontology in percentage.
- **non-cat slot span**: how many values of non-categorical slots have span annotation in percentage.
### Citation
```
@InProceedings{li2017dailydialog,
author = {Li, Yanran and Su, Hui and Shen, Xiaoyu and Li, Wenjie and Cao, Ziqiang and Niu, Shuzi},
title = {DailyDialog: A Manually Labelled Multi-turn Dialogue Dataset},
booktitle = {Proceedings of The 8th International Joint Conference on Natural Language Processing (IJCNLP 2017)},
year = {2017}
}
```
### Licensing Information
[**CC BY-NC-SA 4.0**](https://creativecommons.org/licenses/by-nc-sa/4.0/) |
false |
# Dataset Card for `reviews_with_drift`
## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
### Dataset Summary
This dataset was crafted to be used in our tutorial [Link to the tutorial when ready]. It consists on a large Movie Review Dataset mixed with some reviews from a Hotel Review Dataset. The training/validation set are purely obtained from the Movie Review Dataset while the production set is mixed. Some other features have been added (`age`, `gender`, `context`) as well as a made up timestamp `prediction_ts` of when the inference took place.
### Supported Tasks and Leaderboards
`text-classification`, `sentiment-classification`: The dataset is mainly used for text classification: given the text, predict the sentiment (positive or negative).
### Languages
Text is mainly written in english.
## Dataset Structure
### Data Instances
[More Information Needed]
### Data Fields
[More Information Needed]
### Data Splits
[More Information Needed]
## Dataset Creation
### Curation Rationale
[More Information Needed]
### Source Data
[More Information Needed]
#### Initial Data Collection and Normalization
[More Information Needed]
#### Who are the source language producers?
[More Information Needed]
### Annotations
[More Information Needed]
#### Annotation process
[More Information Needed]
#### Who are the annotators?
[More Information Needed]
### Personal and Sensitive Information
[More Information Needed]
## Considerations for Using the Data
### Social Impact of Dataset
[More Information Needed]
### Discussion of Biases
[More Information Needed]
### Other Known Limitations
[More Information Needed]
## Additional Information
### Dataset Curators
[More Information Needed]
### Licensing Information
[More Information Needed]
### Citation Information
[More Information Needed]
### Contributions
Thanks to [@fjcasti1](https://github.com/fjcasti1) for adding this dataset. |
true |
# Dataset Card for "UnpredicTable-full" - Dataset of Few-shot Tasks from Tables
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-instances)
- [Data Splits](#data-instances)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
## Dataset Description
- **Homepage:** https://ethanperez.net/unpredictable
- **Repository:** https://github.com/JunShern/few-shot-adaptation
- **Paper:** Few-shot Adaptation Works with UnpredicTable Data
- **Point of Contact:** junshern@nyu.edu, perez@nyu.edu
### Dataset Summary
The UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.
There are several dataset versions available:
* [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full): Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full), which comprises 413,299 tasks from 23,744 unique websites.
* [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique): This is the same as [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full) but filtered to have a maximum of one task per website. [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique) contains exactly 23,744 tasks from 23,744 websites.
* [UnpredicTable-5k](https://huggingface.co/datasets/MicPie/unpredictable_5k): This dataset contains 5k random tables from the full dataset.
* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):
* [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low)
* [UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium)
* [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high)
* UnpredicTable data subsets based on the website of origin:
* [UnpredicTable-baseball-fantasysports-yahoo-com](https://huggingface.co/datasets/MicPie/unpredictable_baseball-fantasysports-yahoo-com)
* [UnpredicTable-bulbapedia-bulbagarden-net](https://huggingface.co/datasets/MicPie/unpredictable_bulbapedia-bulbagarden-net)
* [UnpredicTable-cappex-com](https://huggingface.co/datasets/MicPie/unpredictable_cappex-com)
* [UnpredicTable-cram-com](https://huggingface.co/datasets/MicPie/unpredictable_cram-com)
* [UnpredicTable-dividend-com](https://huggingface.co/datasets/MicPie/unpredictable_dividend-com)
* [UnpredicTable-dummies-com](https://huggingface.co/datasets/MicPie/unpredictable_dummies-com)
* [UnpredicTable-en-wikipedia-org](https://huggingface.co/datasets/MicPie/unpredictable_en-wikipedia-org)
* [UnpredicTable-ensembl-org](https://huggingface.co/datasets/MicPie/unpredictable_ensembl-org)
* [UnpredicTable-gamefaqs-com](https://huggingface.co/datasets/MicPie/unpredictable_gamefaqs-com)
* [UnpredicTable-mgoblog-com](https://huggingface.co/datasets/MicPie/unpredictable_mgoblog-com)
* [UnpredicTable-mmo-champion-com](https://huggingface.co/datasets/MicPie/unpredictable_mmo-champion-com)
* [UnpredicTable-msdn-microsoft-com](https://huggingface.co/datasets/MicPie/unpredictable_msdn-microsoft-com)
* [UnpredicTable-phonearena-com](https://huggingface.co/datasets/MicPie/unpredictable_phonearena-com)
* [UnpredicTable-sittercity-com](https://huggingface.co/datasets/MicPie/unpredictable_sittercity-com)
* [UnpredicTable-sporcle-com](https://huggingface.co/datasets/MicPie/unpredictable_sporcle-com)
* [UnpredicTable-studystack-com](https://huggingface.co/datasets/MicPie/unpredictable_studystack-com)
* [UnpredicTable-support-google-com](https://huggingface.co/datasets/MicPie/unpredictable_support-google-com)
* [UnpredicTable-w3-org](https://huggingface.co/datasets/MicPie/unpredictable_w3-org)
* [UnpredicTable-wiki-openmoko-org](https://huggingface.co/datasets/MicPie/unpredictable_wiki-openmoko-org)
* [UnpredicTable-wkdu-org](https://huggingface.co/datasets/MicPie/unpredictable_wkdu-org)
* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):
* [UnpredicTable-cluster00](https://huggingface.co/datasets/MicPie/unpredictable_cluster00)
* [UnpredicTable-cluster01](https://huggingface.co/datasets/MicPie/unpredictable_cluster01)
* [UnpredicTable-cluster02](https://huggingface.co/datasets/MicPie/unpredictable_cluster02)
* [UnpredicTable-cluster03](https://huggingface.co/datasets/MicPie/unpredictable_cluster03)
* [UnpredicTable-cluster04](https://huggingface.co/datasets/MicPie/unpredictable_cluster04)
* [UnpredicTable-cluster05](https://huggingface.co/datasets/MicPie/unpredictable_cluster05)
* [UnpredicTable-cluster06](https://huggingface.co/datasets/MicPie/unpredictable_cluster06)
* [UnpredicTable-cluster07](https://huggingface.co/datasets/MicPie/unpredictable_cluster07)
* [UnpredicTable-cluster08](https://huggingface.co/datasets/MicPie/unpredictable_cluster08)
* [UnpredicTable-cluster09](https://huggingface.co/datasets/MicPie/unpredictable_cluster09)
* [UnpredicTable-cluster10](https://huggingface.co/datasets/MicPie/unpredictable_cluster10)
* [UnpredicTable-cluster11](https://huggingface.co/datasets/MicPie/unpredictable_cluster11)
* [UnpredicTable-cluster12](https://huggingface.co/datasets/MicPie/unpredictable_cluster12)
* [UnpredicTable-cluster13](https://huggingface.co/datasets/MicPie/unpredictable_cluster13)
* [UnpredicTable-cluster14](https://huggingface.co/datasets/MicPie/unpredictable_cluster14)
* [UnpredicTable-cluster15](https://huggingface.co/datasets/MicPie/unpredictable_cluster15)
* [UnpredicTable-cluster16](https://huggingface.co/datasets/MicPie/unpredictable_cluster16)
* [UnpredicTable-cluster17](https://huggingface.co/datasets/MicPie/unpredictable_cluster17)
* [UnpredicTable-cluster18](https://huggingface.co/datasets/MicPie/unpredictable_cluster18)
* [UnpredicTable-cluster19](https://huggingface.co/datasets/MicPie/unpredictable_cluster19)
* [UnpredicTable-cluster20](https://huggingface.co/datasets/MicPie/unpredictable_cluster20)
* [UnpredicTable-cluster21](https://huggingface.co/datasets/MicPie/unpredictable_cluster21)
* [UnpredicTable-cluster22](https://huggingface.co/datasets/MicPie/unpredictable_cluster22)
* [UnpredicTable-cluster23](https://huggingface.co/datasets/MicPie/unpredictable_cluster23)
* [UnpredicTable-cluster24](https://huggingface.co/datasets/MicPie/unpredictable_cluster24)
* [UnpredicTable-cluster25](https://huggingface.co/datasets/MicPie/unpredictable_cluster25)
* [UnpredicTable-cluster26](https://huggingface.co/datasets/MicPie/unpredictable_cluster26)
* [UnpredicTable-cluster27](https://huggingface.co/datasets/MicPie/unpredictable_cluster27)
* [UnpredicTable-cluster28](https://huggingface.co/datasets/MicPie/unpredictable_cluster28)
* [UnpredicTable-cluster29](https://huggingface.co/datasets/MicPie/unpredictable_cluster29)
* [UnpredicTable-cluster-noise](https://huggingface.co/datasets/MicPie/unpredictable_cluster-noise)
### Supported Tasks and Leaderboards
Since the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.
The intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.
### Languages
English
## Dataset Structure
### Data Instances
Each task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.
There are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.
### Data Fields
'task': task identifier
'input': column elements of a specific row in the table.
'options': for multiple choice classification, it provides the options to choose from.
'output': target column element of the same row as input.
'pageTitle': the title of the page containing the table.
'outputColName': output column name
'url': url to the website containing the table
'wdcFile': WDC Web Table Corpus file
### Data Splits
The UnpredicTable datasets do not come with additional data splits.
## Dataset Creation
### Curation Rationale
Few-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.
### Source Data
#### Initial Data Collection and Normalization
We use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (http://webdatacommons.org/webtables/2015/EnglishStatistics.html). The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.
#### Who are the source language producers?
The dataset is extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/).
### Annotations
#### Annotation process
Manual annotation was only carried out for the [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low),
[UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium), and [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high) data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.
#### Who are the annotators?
Annotations were carried out by a lab assistant.
### Personal and Sensitive Information
The data was extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/), which in turn extracted tables from the [Common Crawl](https://commoncrawl.org/). We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.
## Considerations for Using the Data
### Social Impact of Dataset
This dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.
### Discussion of Biases
Since our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.
### Other Known Limitations
No additional known limitations.
## Additional Information
### Dataset Curators
Jun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez
### Licensing Information
Apache 2.0
### Citation Information
```
@misc{chan2022few,
author = {Chan, Jun Shern and Pieler, Michael and Jao, Jonathan and Scheurer, Jérémy and Perez, Ethan},
title = {Few-shot Adaptation Works with UnpredicTable Data},
publisher={arXiv},
year = {2022},
url = {https://arxiv.org/abs/2208.01009}
}
```
|
true |
# Dataset Card for "UnpredicTable-dividend-com" - Dataset of Few-shot Tasks from Tables
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-instances)
- [Data Splits](#data-instances)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
## Dataset Description
- **Homepage:** https://ethanperez.net/unpredictable
- **Repository:** https://github.com/JunShern/few-shot-adaptation
- **Paper:** Few-shot Adaptation Works with UnpredicTable Data
- **Point of Contact:** junshern@nyu.edu, perez@nyu.edu
### Dataset Summary
The UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.
There are several dataset versions available:
* [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full): Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full), which comprises 413,299 tasks from 23,744 unique websites.
* [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique): This is the same as [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full) but filtered to have a maximum of one task per website. [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique) contains exactly 23,744 tasks from 23,744 websites.
* [UnpredicTable-5k](https://huggingface.co/datasets/MicPie/unpredictable_5k): This dataset contains 5k random tables from the full dataset.
* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):
* [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low)
* [UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium)
* [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high)
* UnpredicTable data subsets based on the website of origin:
* [UnpredicTable-baseball-fantasysports-yahoo-com](https://huggingface.co/datasets/MicPie/unpredictable_baseball-fantasysports-yahoo-com)
* [UnpredicTable-bulbapedia-bulbagarden-net](https://huggingface.co/datasets/MicPie/unpredictable_bulbapedia-bulbagarden-net)
* [UnpredicTable-cappex-com](https://huggingface.co/datasets/MicPie/unpredictable_cappex-com)
* [UnpredicTable-cram-com](https://huggingface.co/datasets/MicPie/unpredictable_cram-com)
* [UnpredicTable-dividend-com](https://huggingface.co/datasets/MicPie/unpredictable_dividend-com)
* [UnpredicTable-dummies-com](https://huggingface.co/datasets/MicPie/unpredictable_dummies-com)
* [UnpredicTable-en-wikipedia-org](https://huggingface.co/datasets/MicPie/unpredictable_en-wikipedia-org)
* [UnpredicTable-ensembl-org](https://huggingface.co/datasets/MicPie/unpredictable_ensembl-org)
* [UnpredicTable-gamefaqs-com](https://huggingface.co/datasets/MicPie/unpredictable_gamefaqs-com)
* [UnpredicTable-mgoblog-com](https://huggingface.co/datasets/MicPie/unpredictable_mgoblog-com)
* [UnpredicTable-mmo-champion-com](https://huggingface.co/datasets/MicPie/unpredictable_mmo-champion-com)
* [UnpredicTable-msdn-microsoft-com](https://huggingface.co/datasets/MicPie/unpredictable_msdn-microsoft-com)
* [UnpredicTable-phonearena-com](https://huggingface.co/datasets/MicPie/unpredictable_phonearena-com)
* [UnpredicTable-sittercity-com](https://huggingface.co/datasets/MicPie/unpredictable_sittercity-com)
* [UnpredicTable-sporcle-com](https://huggingface.co/datasets/MicPie/unpredictable_sporcle-com)
* [UnpredicTable-studystack-com](https://huggingface.co/datasets/MicPie/unpredictable_studystack-com)
* [UnpredicTable-support-google-com](https://huggingface.co/datasets/MicPie/unpredictable_support-google-com)
* [UnpredicTable-w3-org](https://huggingface.co/datasets/MicPie/unpredictable_w3-org)
* [UnpredicTable-wiki-openmoko-org](https://huggingface.co/datasets/MicPie/unpredictable_wiki-openmoko-org)
* [UnpredicTable-wkdu-org](https://huggingface.co/datasets/MicPie/unpredictable_wkdu-org)
* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):
* [UnpredicTable-cluster00](https://huggingface.co/datasets/MicPie/unpredictable_cluster00)
* [UnpredicTable-cluster01](https://huggingface.co/datasets/MicPie/unpredictable_cluster01)
* [UnpredicTable-cluster02](https://huggingface.co/datasets/MicPie/unpredictable_cluster02)
* [UnpredicTable-cluster03](https://huggingface.co/datasets/MicPie/unpredictable_cluster03)
* [UnpredicTable-cluster04](https://huggingface.co/datasets/MicPie/unpredictable_cluster04)
* [UnpredicTable-cluster05](https://huggingface.co/datasets/MicPie/unpredictable_cluster05)
* [UnpredicTable-cluster06](https://huggingface.co/datasets/MicPie/unpredictable_cluster06)
* [UnpredicTable-cluster07](https://huggingface.co/datasets/MicPie/unpredictable_cluster07)
* [UnpredicTable-cluster08](https://huggingface.co/datasets/MicPie/unpredictable_cluster08)
* [UnpredicTable-cluster09](https://huggingface.co/datasets/MicPie/unpredictable_cluster09)
* [UnpredicTable-cluster10](https://huggingface.co/datasets/MicPie/unpredictable_cluster10)
* [UnpredicTable-cluster11](https://huggingface.co/datasets/MicPie/unpredictable_cluster11)
* [UnpredicTable-cluster12](https://huggingface.co/datasets/MicPie/unpredictable_cluster12)
* [UnpredicTable-cluster13](https://huggingface.co/datasets/MicPie/unpredictable_cluster13)
* [UnpredicTable-cluster14](https://huggingface.co/datasets/MicPie/unpredictable_cluster14)
* [UnpredicTable-cluster15](https://huggingface.co/datasets/MicPie/unpredictable_cluster15)
* [UnpredicTable-cluster16](https://huggingface.co/datasets/MicPie/unpredictable_cluster16)
* [UnpredicTable-cluster17](https://huggingface.co/datasets/MicPie/unpredictable_cluster17)
* [UnpredicTable-cluster18](https://huggingface.co/datasets/MicPie/unpredictable_cluster18)
* [UnpredicTable-cluster19](https://huggingface.co/datasets/MicPie/unpredictable_cluster19)
* [UnpredicTable-cluster20](https://huggingface.co/datasets/MicPie/unpredictable_cluster20)
* [UnpredicTable-cluster21](https://huggingface.co/datasets/MicPie/unpredictable_cluster21)
* [UnpredicTable-cluster22](https://huggingface.co/datasets/MicPie/unpredictable_cluster22)
* [UnpredicTable-cluster23](https://huggingface.co/datasets/MicPie/unpredictable_cluster23)
* [UnpredicTable-cluster24](https://huggingface.co/datasets/MicPie/unpredictable_cluster24)
* [UnpredicTable-cluster25](https://huggingface.co/datasets/MicPie/unpredictable_cluster25)
* [UnpredicTable-cluster26](https://huggingface.co/datasets/MicPie/unpredictable_cluster26)
* [UnpredicTable-cluster27](https://huggingface.co/datasets/MicPie/unpredictable_cluster27)
* [UnpredicTable-cluster28](https://huggingface.co/datasets/MicPie/unpredictable_cluster28)
* [UnpredicTable-cluster29](https://huggingface.co/datasets/MicPie/unpredictable_cluster29)
* [UnpredicTable-cluster-noise](https://huggingface.co/datasets/MicPie/unpredictable_cluster-noise)
### Supported Tasks and Leaderboards
Since the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.
The intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.
### Languages
English
## Dataset Structure
### Data Instances
Each task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.
There are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.
### Data Fields
'task': task identifier
'input': column elements of a specific row in the table.
'options': for multiple choice classification, it provides the options to choose from.
'output': target column element of the same row as input.
'pageTitle': the title of the page containing the table.
'outputColName': output column name
'url': url to the website containing the table
'wdcFile': WDC Web Table Corpus file
### Data Splits
The UnpredicTable datasets do not come with additional data splits.
## Dataset Creation
### Curation Rationale
Few-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.
### Source Data
#### Initial Data Collection and Normalization
We use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (http://webdatacommons.org/webtables/2015/EnglishStatistics.html). The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.
#### Who are the source language producers?
The dataset is extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/).
### Annotations
#### Annotation process
Manual annotation was only carried out for the [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low),
[UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium), and [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high) data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.
#### Who are the annotators?
Annotations were carried out by a lab assistant.
### Personal and Sensitive Information
The data was extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/), which in turn extracted tables from the [Common Crawl](https://commoncrawl.org/). We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.
## Considerations for Using the Data
### Social Impact of Dataset
This dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.
### Discussion of Biases
Since our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.
### Other Known Limitations
No additional known limitations.
## Additional Information
### Dataset Curators
Jun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez
### Licensing Information
Apache 2.0
### Citation Information
```
@misc{chan2022few,
author = {Chan, Jun Shern and Pieler, Michael and Jao, Jonathan and Scheurer, Jérémy and Perez, Ethan},
title = {Few-shot Adaptation Works with UnpredicTable Data},
publisher={arXiv},
year = {2022},
url = {https://arxiv.org/abs/2208.01009}
}
```
|
true |
# Dataset Card for "UnpredicTable-studystack-com" - Dataset of Few-shot Tasks from Tables
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-instances)
- [Data Splits](#data-instances)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
## Dataset Description
- **Homepage:** https://ethanperez.net/unpredictable
- **Repository:** https://github.com/JunShern/few-shot-adaptation
- **Paper:** Few-shot Adaptation Works with UnpredicTable Data
- **Point of Contact:** junshern@nyu.edu, perez@nyu.edu
### Dataset Summary
The UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.
There are several dataset versions available:
* [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full): Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full), which comprises 413,299 tasks from 23,744 unique websites.
* [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique): This is the same as [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full) but filtered to have a maximum of one task per website. [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique) contains exactly 23,744 tasks from 23,744 websites.
* [UnpredicTable-5k](https://huggingface.co/datasets/MicPie/unpredictable_5k): This dataset contains 5k random tables from the full dataset.
* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):
* [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low)
* [UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium)
* [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high)
* UnpredicTable data subsets based on the website of origin:
* [UnpredicTable-baseball-fantasysports-yahoo-com](https://huggingface.co/datasets/MicPie/unpredictable_baseball-fantasysports-yahoo-com)
* [UnpredicTable-bulbapedia-bulbagarden-net](https://huggingface.co/datasets/MicPie/unpredictable_bulbapedia-bulbagarden-net)
* [UnpredicTable-cappex-com](https://huggingface.co/datasets/MicPie/unpredictable_cappex-com)
* [UnpredicTable-cram-com](https://huggingface.co/datasets/MicPie/unpredictable_cram-com)
* [UnpredicTable-dividend-com](https://huggingface.co/datasets/MicPie/unpredictable_dividend-com)
* [UnpredicTable-dummies-com](https://huggingface.co/datasets/MicPie/unpredictable_dummies-com)
* [UnpredicTable-en-wikipedia-org](https://huggingface.co/datasets/MicPie/unpredictable_en-wikipedia-org)
* [UnpredicTable-ensembl-org](https://huggingface.co/datasets/MicPie/unpredictable_ensembl-org)
* [UnpredicTable-gamefaqs-com](https://huggingface.co/datasets/MicPie/unpredictable_gamefaqs-com)
* [UnpredicTable-mgoblog-com](https://huggingface.co/datasets/MicPie/unpredictable_mgoblog-com)
* [UnpredicTable-mmo-champion-com](https://huggingface.co/datasets/MicPie/unpredictable_mmo-champion-com)
* [UnpredicTable-msdn-microsoft-com](https://huggingface.co/datasets/MicPie/unpredictable_msdn-microsoft-com)
* [UnpredicTable-phonearena-com](https://huggingface.co/datasets/MicPie/unpredictable_phonearena-com)
* [UnpredicTable-sittercity-com](https://huggingface.co/datasets/MicPie/unpredictable_sittercity-com)
* [UnpredicTable-sporcle-com](https://huggingface.co/datasets/MicPie/unpredictable_sporcle-com)
* [UnpredicTable-studystack-com](https://huggingface.co/datasets/MicPie/unpredictable_studystack-com)
* [UnpredicTable-support-google-com](https://huggingface.co/datasets/MicPie/unpredictable_support-google-com)
* [UnpredicTable-w3-org](https://huggingface.co/datasets/MicPie/unpredictable_w3-org)
* [UnpredicTable-wiki-openmoko-org](https://huggingface.co/datasets/MicPie/unpredictable_wiki-openmoko-org)
* [UnpredicTable-wkdu-org](https://huggingface.co/datasets/MicPie/unpredictable_wkdu-org)
* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):
* [UnpredicTable-cluster00](https://huggingface.co/datasets/MicPie/unpredictable_cluster00)
* [UnpredicTable-cluster01](https://huggingface.co/datasets/MicPie/unpredictable_cluster01)
* [UnpredicTable-cluster02](https://huggingface.co/datasets/MicPie/unpredictable_cluster02)
* [UnpredicTable-cluster03](https://huggingface.co/datasets/MicPie/unpredictable_cluster03)
* [UnpredicTable-cluster04](https://huggingface.co/datasets/MicPie/unpredictable_cluster04)
* [UnpredicTable-cluster05](https://huggingface.co/datasets/MicPie/unpredictable_cluster05)
* [UnpredicTable-cluster06](https://huggingface.co/datasets/MicPie/unpredictable_cluster06)
* [UnpredicTable-cluster07](https://huggingface.co/datasets/MicPie/unpredictable_cluster07)
* [UnpredicTable-cluster08](https://huggingface.co/datasets/MicPie/unpredictable_cluster08)
* [UnpredicTable-cluster09](https://huggingface.co/datasets/MicPie/unpredictable_cluster09)
* [UnpredicTable-cluster10](https://huggingface.co/datasets/MicPie/unpredictable_cluster10)
* [UnpredicTable-cluster11](https://huggingface.co/datasets/MicPie/unpredictable_cluster11)
* [UnpredicTable-cluster12](https://huggingface.co/datasets/MicPie/unpredictable_cluster12)
* [UnpredicTable-cluster13](https://huggingface.co/datasets/MicPie/unpredictable_cluster13)
* [UnpredicTable-cluster14](https://huggingface.co/datasets/MicPie/unpredictable_cluster14)
* [UnpredicTable-cluster15](https://huggingface.co/datasets/MicPie/unpredictable_cluster15)
* [UnpredicTable-cluster16](https://huggingface.co/datasets/MicPie/unpredictable_cluster16)
* [UnpredicTable-cluster17](https://huggingface.co/datasets/MicPie/unpredictable_cluster17)
* [UnpredicTable-cluster18](https://huggingface.co/datasets/MicPie/unpredictable_cluster18)
* [UnpredicTable-cluster19](https://huggingface.co/datasets/MicPie/unpredictable_cluster19)
* [UnpredicTable-cluster20](https://huggingface.co/datasets/MicPie/unpredictable_cluster20)
* [UnpredicTable-cluster21](https://huggingface.co/datasets/MicPie/unpredictable_cluster21)
* [UnpredicTable-cluster22](https://huggingface.co/datasets/MicPie/unpredictable_cluster22)
* [UnpredicTable-cluster23](https://huggingface.co/datasets/MicPie/unpredictable_cluster23)
* [UnpredicTable-cluster24](https://huggingface.co/datasets/MicPie/unpredictable_cluster24)
* [UnpredicTable-cluster25](https://huggingface.co/datasets/MicPie/unpredictable_cluster25)
* [UnpredicTable-cluster26](https://huggingface.co/datasets/MicPie/unpredictable_cluster26)
* [UnpredicTable-cluster27](https://huggingface.co/datasets/MicPie/unpredictable_cluster27)
* [UnpredicTable-cluster28](https://huggingface.co/datasets/MicPie/unpredictable_cluster28)
* [UnpredicTable-cluster29](https://huggingface.co/datasets/MicPie/unpredictable_cluster29)
* [UnpredicTable-cluster-noise](https://huggingface.co/datasets/MicPie/unpredictable_cluster-noise)
### Supported Tasks and Leaderboards
Since the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.
The intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.
### Languages
English
## Dataset Structure
### Data Instances
Each task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.
There are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.
### Data Fields
'task': task identifier
'input': column elements of a specific row in the table.
'options': for multiple choice classification, it provides the options to choose from.
'output': target column element of the same row as input.
'pageTitle': the title of the page containing the table.
'outputColName': output column name
'url': url to the website containing the table
'wdcFile': WDC Web Table Corpus file
### Data Splits
The UnpredicTable datasets do not come with additional data splits.
## Dataset Creation
### Curation Rationale
Few-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.
### Source Data
#### Initial Data Collection and Normalization
We use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (http://webdatacommons.org/webtables/2015/EnglishStatistics.html). The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.
#### Who are the source language producers?
The dataset is extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/).
### Annotations
#### Annotation process
Manual annotation was only carried out for the [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low),
[UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium), and [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high) data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.
#### Who are the annotators?
Annotations were carried out by a lab assistant.
### Personal and Sensitive Information
The data was extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/), which in turn extracted tables from the [Common Crawl](https://commoncrawl.org/). We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.
## Considerations for Using the Data
### Social Impact of Dataset
This dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.
### Discussion of Biases
Since our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.
### Other Known Limitations
No additional known limitations.
## Additional Information
### Dataset Curators
Jun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez
### Licensing Information
Apache 2.0
### Citation Information
```
@misc{chan2022few,
author = {Chan, Jun Shern and Pieler, Michael and Jao, Jonathan and Scheurer, Jérémy and Perez, Ethan},
title = {Few-shot Adaptation Works with UnpredicTable Data},
publisher={arXiv},
year = {2022},
url = {https://arxiv.org/abs/2208.01009}
}
```
|
true | # Dataset Card for Multilingual HateCheck
## Dataset Description
Multilingual HateCheck (MHC) is a suite of functional tests for hate speech detection models in 10 different languages: Arabic, Dutch, French, German, Hindi, Italian, Mandarin, Polish, Portuguese and Spanish.
For each language, there are 25+ functional tests that correspond to distinct types of hate and challenging non-hate.
This allows for targeted diagnostic insights into model performance.
For more details, please refer to our paper about MHC, published at the 2022 Workshop on Online Abuse and Harms (WOAH) at NAACL 2022. If you are using MHC, please cite our work!
- **Paper:** Röttger et al. (2022) - Multilingual HateCheck: Functional Tests for Multilingual Hate Speech Detection Models. https://arxiv.org/abs/2206.09917
- **Repository:** https://github.com/rewire-online/multilingual-hatecheck
- **Point of Contact:** paul@rewire.online
## Dataset Structure
The csv format mostly matches the original HateCheck data, with some adjustments for specific languages.
**mhc_case_id**
The test case ID that is unique to each test case across languages (e.g., "mandarin-1305")
**functionality**
The shorthand for the functionality tested by the test case (e.g, "target_obj_nh"). The same functionalities are tested in all languages, except for Mandarin and Arabic, where non-Latin script required adapting the tests for spelling variations.
**test_case**
The test case text.
**label_gold**
The gold standard label ("hateful" or "non-hateful") of the test case. All test cases within a given functionality have the same gold standard label.
**target_ident**
Where applicable, the protected group that is targeted or referenced in the test case. All HateChecks cover seven target groups, but their composition varies across languages.
**ref_case_id**
For hateful cases, where applicable, the ID of the hateful case which was perturbed to generate this test case. For non-hateful cases, where applicable, the ID of the hateful case which is contrasted by this test case.
**ref_templ_id**
The equivalent to ref_case_id, but for template IDs.
**templ_id**
The ID of the template from which the test case was generated.
**case_templ**
The template from which the test case was generated (where applicable).
**gender_male** and **gender_female**
For gender-inflected languages (French, Spanish, Portuguese, Hindi, Arabic, Italian, Polish, German), only for cases where gender inflection is relevant, separate entries for gender_male and gender_female replace case_templ.
**label_annotated**
A list of labels given by the three annotators who reviewed the test case (e.g., "['hateful', 'hateful', 'hateful']").
**label_annotated_maj**
The majority vote of the three annotators (e.g., "hateful"). In some cases this differs from the gold label given by our language experts.
**disagreement_in_case**
True if label_annotated_maj does not match label_gold for the entry.
**disagreement_in_template**
True if the test case is generated from an IDENT template and there is at least one case with disagreement_in_case generated from the same template. This can be used to exclude entire templates from MHC. |
false |
# Dataset Card for Natural Questions
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **Homepage:** [https://ai.google.com/research/NaturalQuestions/dataset](https://ai.google.com/research/NaturalQuestions/dataset)
- **Repository:** [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
- **Paper:** [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
- **Point of Contact:** [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
- **Size of downloaded dataset files:** 42981 MB
- **Size of the generated dataset:** 139706 MB
- **Total amount of disk used:** 182687 MB
### Dataset Summary
The NQ corpus contains questions from real users, and it requires QA systems to
read and comprehend an entire Wikipedia article that may or may not contain the
answer to the question. The inclusion of real user questions, and the
requirement that solutions should read an entire page to find the answer, cause
NQ to be a more realistic and challenging task than prior QA datasets.
### Supported Tasks and Leaderboards
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Languages
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
## Dataset Structure
### Data Instances
#### default
- **Size of downloaded dataset files:** 42981 MB
- **Size of the generated dataset:** 139706 MB
- **Total amount of disk used:** 182687 MB
An example of 'train' looks as follows.
```
```
### Data Fields
The data fields are the same among all splits.
#### default
```
"id": datasets.Value("string"),
"document": {
"title": datasets.Value("string"),
"url": datasets.Value("string"),
"html": datasets.Value("string"),
"tokens": datasets.features.Sequence(
{
"token": datasets.Value("string"),
"is_html": datasets.Value("bool"),
"start_byte": datasets.Value("int64"),
"end_byte": datasets.Value("int64"),
}
),
},
"question": {
"text": datasets.Value("string"),
"tokens": datasets.features.Sequence(datasets.Value("string")),
},
"long_answer_candidates": datasets.features.Sequence(
{
"start_token": datasets.Value("int64"),
"end_token": datasets.Value("int64"),
"start_byte": datasets.Value("int64"),
"end_byte": datasets.Value("int64"),
"top_level": datasets.Value("bool"),
}
),
"annotations": datasets.features.Sequence(
{
"id": datasets.Value("string"),
"long_answer": {
"start_token": datasets.Value("int64"),
"end_token": datasets.Value("int64"),
"start_byte": datasets.Value("int64"),
"end_byte": datasets.Value("int64"),
"candidate_index": datasets.Value("int64")
},
"short_answers": datasets.features.Sequence(
{
"start_token": datasets.Value("int64"),
"end_token": datasets.Value("int64"),
"start_byte": datasets.Value("int64"),
"end_byte": datasets.Value("int64"),
"text": datasets.Value("string"),
}
),
"yes_no_answer": datasets.features.ClassLabel(
names=["NO", "YES"]
), # Can also be -1 for NONE.
}
)
```
### Data Splits
| name | train | validation |
|---------|-------:|-----------:|
| default | 307373 | 7830 |
| dev | N/A | 7830 |
## Dataset Creation
### Curation Rationale
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Source Data
#### Initial Data Collection and Normalization
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
#### Who are the source language producers?
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Annotations
#### Annotation process
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
#### Who are the annotators?
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Personal and Sensitive Information
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
## Considerations for Using the Data
### Social Impact of Dataset
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Discussion of Biases
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Other Known Limitations
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
## Additional Information
### Dataset Curators
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Licensing Information
[Creative Commons Attribution-ShareAlike 3.0 Unported](https://creativecommons.org/licenses/by-sa/3.0/).
### Citation Information
```
@article{47761,
title = {Natural Questions: a Benchmark for Question Answering Research},
author = {Tom Kwiatkowski and Jennimaria Palomaki and Olivia Redfield and Michael Collins and Ankur Parikh and Chris Alberti and Danielle Epstein and Illia Polosukhin and Matthew Kelcey and Jacob Devlin and Kenton Lee and Kristina N. Toutanova and Llion Jones and Ming-Wei Chang and Andrew Dai and Jakob Uszkoreit and Quoc Le and Slav Petrov},
year = {2019},
journal = {Transactions of the Association of Computational Linguistics}
}
```
### Contributions
|
false |
# Dataset Card for ogbg-ppa
## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [External Use](#external-use)
- [PyGeometric](#pygeometric)
- [Dataset Structure](#dataset-structure)
- [Data Properties](#data-properties)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Additional Information](#additional-information)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
- **[Homepage](https://ogb.stanford.edu/docs/graphprop/#ogbg-ppa)**
- **[Repository](https://github.com/snap-stanford/ogb):**:
- **Paper:**: Open Graph Benchmark: Datasets for Machine Learning on Graphs (see citation)
- **Leaderboard:**: [OGB leaderboard](https://ogb.stanford.edu/docs/leader_graphprop/#ogbg-ppa) and [Papers with code leaderboard](https://paperswithcode.com/sota/graph-property-prediction-on-ogbg-ppa)
### Dataset Summary
The `ogbg-ppa` dataset is "a set of undirected protein association neighborhoods extracted from the protein-protein association networks of 1,581 species", over 37 taxonomic groups, by teams at Stanford, to be a part of the Open Graph Benchmark. See their website for dataset postprocessing.
### Supported Tasks and Leaderboards
`ogbg-ppa` should be used for taxonomic group prediction, a 37-way multi-class classification task. The score used is Average Precision on the test set.
## External Use
### PyGeometric
To load in PyGeometric, do the following:
```python
from datasets import load_dataset
from torch_geometric.data import Data
from torch_geometric.loader import DataLoader
graphs_dataset = load_dataset("graphs-datasets/ogbg-ppa")
# For the train set (replace by valid or test as needed)
graphs_list = [Data(graph) for graph in graphs_dataset["train"]]
graphs_pygeometric = DataLoader(graph_list)
```
## Dataset Structure
### Data Properties
| property | value |
|---|---|
| scale | small |
| #graphs | 158,100 |
| average #nodes | 243.4 |
| average #edges | 2,266.1 |
| average node degree | 18.3 |
| average cluster coefficient | 0.513 |
| MaxSCC ratio | 1.000 |
| graph diameter | 4.8 |
### Data Fields
Each row of a given file is a graph, with:
- `edge_index` (list: 2 x #edges): pairs of nodes constituting edges
- `edge_attr` (list: #edges x #edge-features): for the aforementioned edges, contains their features
- `y` (list: 1 x #labels): contains the number of labels available to predict (here 1, equal to zero or one)
- `num_nodes` (int): number of nodes of the graph
The nodes don't have specific features and are implicit from the lists of edges
### Data Splits
This data comes from the PyGeometric version of the dataset provided by OGB, and follows the provided data splits.
This information can be found back using
```python
from ogb.graphproppred import PygGraphPropPredDataset
dataset = PygGraphPropPredDataset(name = 'ogbg-ppa')
split_idx = dataset.get_idx_split()
train = dataset[split_idx['train']] # valid, test
```
## Additional Information
### Licensing Information
The dataset has been released under CC-0 license.
### Citation Information
```
@inproceedings{hu-etal-2020-open,
author = {Weihua Hu and
Matthias Fey and
Marinka Zitnik and
Yuxiao Dong and
Hongyu Ren and
Bowen Liu and
Michele Catasta and
Jure Leskovec},
editor = {Hugo Larochelle and
Marc Aurelio Ranzato and
Raia Hadsell and
Maria{-}Florina Balcan and
Hsuan{-}Tien Lin},
title = {Open Graph Benchmark: Datasets for Machine Learning on Graphs},
booktitle = {Advances in Neural Information Processing Systems 33: Annual Conference
on Neural Information Processing Systems 2020, NeurIPS 2020, December
6-12, 2020, virtual},
year = {2020},
url = {https://proceedings.neurips.cc/paper/2020/hash/fb60d411a5c5b72b2e7d3527cfc84fd0-Abstract.html},
}
```
### Contributions
Thanks to [@clefourrier](https://github.com/clefourrier) for adding this dataset. |
true |
# Dataset Card for "UnpredicTable-unique" - Dataset of Few-shot Tasks from Tables
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-instances)
- [Data Splits](#data-instances)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
## Dataset Description
- **Homepage:** https://ethanperez.net/unpredictable
- **Repository:** https://github.com/JunShern/few-shot-adaptation
- **Paper:** Few-shot Adaptation Works with UnpredicTable Data
- **Point of Contact:** junshern@nyu.edu, perez@nyu.edu
### Dataset Summary
The UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.
There are several dataset versions available:
* [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full): Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full), which comprises 413,299 tasks from 23,744 unique websites.
* [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique): This is the same as [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full) but filtered to have a maximum of one task per website. [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique) contains exactly 23,744 tasks from 23,744 websites.
* [UnpredicTable-5k](https://huggingface.co/datasets/MicPie/unpredictable_5k): This dataset contains 5k random tables from the full dataset.
* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):
* [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low)
* [UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium)
* [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high)
* UnpredicTable data subsets based on the website of origin:
* [UnpredicTable-baseball-fantasysports-yahoo-com](https://huggingface.co/datasets/MicPie/unpredictable_baseball-fantasysports-yahoo-com)
* [UnpredicTable-bulbapedia-bulbagarden-net](https://huggingface.co/datasets/MicPie/unpredictable_bulbapedia-bulbagarden-net)
* [UnpredicTable-cappex-com](https://huggingface.co/datasets/MicPie/unpredictable_cappex-com)
* [UnpredicTable-cram-com](https://huggingface.co/datasets/MicPie/unpredictable_cram-com)
* [UnpredicTable-dividend-com](https://huggingface.co/datasets/MicPie/unpredictable_dividend-com)
* [UnpredicTable-dummies-com](https://huggingface.co/datasets/MicPie/unpredictable_dummies-com)
* [UnpredicTable-en-wikipedia-org](https://huggingface.co/datasets/MicPie/unpredictable_en-wikipedia-org)
* [UnpredicTable-ensembl-org](https://huggingface.co/datasets/MicPie/unpredictable_ensembl-org)
* [UnpredicTable-gamefaqs-com](https://huggingface.co/datasets/MicPie/unpredictable_gamefaqs-com)
* [UnpredicTable-mgoblog-com](https://huggingface.co/datasets/MicPie/unpredictable_mgoblog-com)
* [UnpredicTable-mmo-champion-com](https://huggingface.co/datasets/MicPie/unpredictable_mmo-champion-com)
* [UnpredicTable-msdn-microsoft-com](https://huggingface.co/datasets/MicPie/unpredictable_msdn-microsoft-com)
* [UnpredicTable-phonearena-com](https://huggingface.co/datasets/MicPie/unpredictable_phonearena-com)
* [UnpredicTable-sittercity-com](https://huggingface.co/datasets/MicPie/unpredictable_sittercity-com)
* [UnpredicTable-sporcle-com](https://huggingface.co/datasets/MicPie/unpredictable_sporcle-com)
* [UnpredicTable-studystack-com](https://huggingface.co/datasets/MicPie/unpredictable_studystack-com)
* [UnpredicTable-support-google-com](https://huggingface.co/datasets/MicPie/unpredictable_support-google-com)
* [UnpredicTable-w3-org](https://huggingface.co/datasets/MicPie/unpredictable_w3-org)
* [UnpredicTable-wiki-openmoko-org](https://huggingface.co/datasets/MicPie/unpredictable_wiki-openmoko-org)
* [UnpredicTable-wkdu-org](https://huggingface.co/datasets/MicPie/unpredictable_wkdu-org)
* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):
* [UnpredicTable-cluster00](https://huggingface.co/datasets/MicPie/unpredictable_cluster00)
* [UnpredicTable-cluster01](https://huggingface.co/datasets/MicPie/unpredictable_cluster01)
* [UnpredicTable-cluster02](https://huggingface.co/datasets/MicPie/unpredictable_cluster02)
* [UnpredicTable-cluster03](https://huggingface.co/datasets/MicPie/unpredictable_cluster03)
* [UnpredicTable-cluster04](https://huggingface.co/datasets/MicPie/unpredictable_cluster04)
* [UnpredicTable-cluster05](https://huggingface.co/datasets/MicPie/unpredictable_cluster05)
* [UnpredicTable-cluster06](https://huggingface.co/datasets/MicPie/unpredictable_cluster06)
* [UnpredicTable-cluster07](https://huggingface.co/datasets/MicPie/unpredictable_cluster07)
* [UnpredicTable-cluster08](https://huggingface.co/datasets/MicPie/unpredictable_cluster08)
* [UnpredicTable-cluster09](https://huggingface.co/datasets/MicPie/unpredictable_cluster09)
* [UnpredicTable-cluster10](https://huggingface.co/datasets/MicPie/unpredictable_cluster10)
* [UnpredicTable-cluster11](https://huggingface.co/datasets/MicPie/unpredictable_cluster11)
* [UnpredicTable-cluster12](https://huggingface.co/datasets/MicPie/unpredictable_cluster12)
* [UnpredicTable-cluster13](https://huggingface.co/datasets/MicPie/unpredictable_cluster13)
* [UnpredicTable-cluster14](https://huggingface.co/datasets/MicPie/unpredictable_cluster14)
* [UnpredicTable-cluster15](https://huggingface.co/datasets/MicPie/unpredictable_cluster15)
* [UnpredicTable-cluster16](https://huggingface.co/datasets/MicPie/unpredictable_cluster16)
* [UnpredicTable-cluster17](https://huggingface.co/datasets/MicPie/unpredictable_cluster17)
* [UnpredicTable-cluster18](https://huggingface.co/datasets/MicPie/unpredictable_cluster18)
* [UnpredicTable-cluster19](https://huggingface.co/datasets/MicPie/unpredictable_cluster19)
* [UnpredicTable-cluster20](https://huggingface.co/datasets/MicPie/unpredictable_cluster20)
* [UnpredicTable-cluster21](https://huggingface.co/datasets/MicPie/unpredictable_cluster21)
* [UnpredicTable-cluster22](https://huggingface.co/datasets/MicPie/unpredictable_cluster22)
* [UnpredicTable-cluster23](https://huggingface.co/datasets/MicPie/unpredictable_cluster23)
* [UnpredicTable-cluster24](https://huggingface.co/datasets/MicPie/unpredictable_cluster24)
* [UnpredicTable-cluster25](https://huggingface.co/datasets/MicPie/unpredictable_cluster25)
* [UnpredicTable-cluster26](https://huggingface.co/datasets/MicPie/unpredictable_cluster26)
* [UnpredicTable-cluster27](https://huggingface.co/datasets/MicPie/unpredictable_cluster27)
* [UnpredicTable-cluster28](https://huggingface.co/datasets/MicPie/unpredictable_cluster28)
* [UnpredicTable-cluster29](https://huggingface.co/datasets/MicPie/unpredictable_cluster29)
* [UnpredicTable-cluster-noise](https://huggingface.co/datasets/MicPie/unpredictable_cluster-noise)
### Supported Tasks and Leaderboards
Since the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.
The intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.
### Languages
English
## Dataset Structure
### Data Instances
Each task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.
There are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.
### Data Fields
'task': task identifier
'input': column elements of a specific row in the table.
'options': for multiple choice classification, it provides the options to choose from.
'output': target column element of the same row as input.
'pageTitle': the title of the page containing the table.
'outputColName': output column name
'url': url to the website containing the table
'wdcFile': WDC Web Table Corpus file
### Data Splits
The UnpredicTable datasets do not come with additional data splits.
## Dataset Creation
### Curation Rationale
Few-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.
### Source Data
#### Initial Data Collection and Normalization
We use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (http://webdatacommons.org/webtables/2015/EnglishStatistics.html). The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.
#### Who are the source language producers?
The dataset is extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/).
### Annotations
#### Annotation process
Manual annotation was only carried out for the [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low),
[UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium), and [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high) data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.
#### Who are the annotators?
Annotations were carried out by a lab assistant.
### Personal and Sensitive Information
The data was extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/), which in turn extracted tables from the [Common Crawl](https://commoncrawl.org/). We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.
## Considerations for Using the Data
### Social Impact of Dataset
This dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.
### Discussion of Biases
Since our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.
### Other Known Limitations
No additional known limitations.
## Additional Information
### Dataset Curators
Jun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez
### Licensing Information
Apache 2.0
### Citation Information
```
@misc{chan2022few,
author = {Chan, Jun Shern and Pieler, Michael and Jao, Jonathan and Scheurer, Jérémy and Perez, Ethan},
title = {Few-shot Adaptation Works with UnpredicTable Data},
publisher={arXiv},
year = {2022},
url = {https://arxiv.org/abs/2208.01009}
}
```
|
true |
# Dataset Card for "UnpredicTable-cluster11" - Dataset of Few-shot Tasks from Tables
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-instances)
- [Data Splits](#data-instances)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
## Dataset Description
- **Homepage:** https://ethanperez.net/unpredictable
- **Repository:** https://github.com/JunShern/few-shot-adaptation
- **Paper:** Few-shot Adaptation Works with UnpredicTable Data
- **Point of Contact:** junshern@nyu.edu, perez@nyu.edu
### Dataset Summary
The UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.
There are several dataset versions available:
* [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full): Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full), which comprises 413,299 tasks from 23,744 unique websites.
* [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique): This is the same as [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full) but filtered to have a maximum of one task per website. [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique) contains exactly 23,744 tasks from 23,744 websites.
* [UnpredicTable-5k](https://huggingface.co/datasets/MicPie/unpredictable_5k): This dataset contains 5k random tables from the full dataset.
* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):
* [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low)
* [UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium)
* [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high)
* UnpredicTable data subsets based on the website of origin:
* [UnpredicTable-baseball-fantasysports-yahoo-com](https://huggingface.co/datasets/MicPie/unpredictable_baseball-fantasysports-yahoo-com)
* [UnpredicTable-bulbapedia-bulbagarden-net](https://huggingface.co/datasets/MicPie/unpredictable_bulbapedia-bulbagarden-net)
* [UnpredicTable-cappex-com](https://huggingface.co/datasets/MicPie/unpredictable_cappex-com)
* [UnpredicTable-cram-com](https://huggingface.co/datasets/MicPie/unpredictable_cram-com)
* [UnpredicTable-dividend-com](https://huggingface.co/datasets/MicPie/unpredictable_dividend-com)
* [UnpredicTable-dummies-com](https://huggingface.co/datasets/MicPie/unpredictable_dummies-com)
* [UnpredicTable-en-wikipedia-org](https://huggingface.co/datasets/MicPie/unpredictable_en-wikipedia-org)
* [UnpredicTable-ensembl-org](https://huggingface.co/datasets/MicPie/unpredictable_ensembl-org)
* [UnpredicTable-gamefaqs-com](https://huggingface.co/datasets/MicPie/unpredictable_gamefaqs-com)
* [UnpredicTable-mgoblog-com](https://huggingface.co/datasets/MicPie/unpredictable_mgoblog-com)
* [UnpredicTable-mmo-champion-com](https://huggingface.co/datasets/MicPie/unpredictable_mmo-champion-com)
* [UnpredicTable-msdn-microsoft-com](https://huggingface.co/datasets/MicPie/unpredictable_msdn-microsoft-com)
* [UnpredicTable-phonearena-com](https://huggingface.co/datasets/MicPie/unpredictable_phonearena-com)
* [UnpredicTable-sittercity-com](https://huggingface.co/datasets/MicPie/unpredictable_sittercity-com)
* [UnpredicTable-sporcle-com](https://huggingface.co/datasets/MicPie/unpredictable_sporcle-com)
* [UnpredicTable-studystack-com](https://huggingface.co/datasets/MicPie/unpredictable_studystack-com)
* [UnpredicTable-support-google-com](https://huggingface.co/datasets/MicPie/unpredictable_support-google-com)
* [UnpredicTable-w3-org](https://huggingface.co/datasets/MicPie/unpredictable_w3-org)
* [UnpredicTable-wiki-openmoko-org](https://huggingface.co/datasets/MicPie/unpredictable_wiki-openmoko-org)
* [UnpredicTable-wkdu-org](https://huggingface.co/datasets/MicPie/unpredictable_wkdu-org)
* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):
* [UnpredicTable-cluster00](https://huggingface.co/datasets/MicPie/unpredictable_cluster00)
* [UnpredicTable-cluster01](https://huggingface.co/datasets/MicPie/unpredictable_cluster01)
* [UnpredicTable-cluster02](https://huggingface.co/datasets/MicPie/unpredictable_cluster02)
* [UnpredicTable-cluster03](https://huggingface.co/datasets/MicPie/unpredictable_cluster03)
* [UnpredicTable-cluster04](https://huggingface.co/datasets/MicPie/unpredictable_cluster04)
* [UnpredicTable-cluster05](https://huggingface.co/datasets/MicPie/unpredictable_cluster05)
* [UnpredicTable-cluster06](https://huggingface.co/datasets/MicPie/unpredictable_cluster06)
* [UnpredicTable-cluster07](https://huggingface.co/datasets/MicPie/unpredictable_cluster07)
* [UnpredicTable-cluster08](https://huggingface.co/datasets/MicPie/unpredictable_cluster08)
* [UnpredicTable-cluster09](https://huggingface.co/datasets/MicPie/unpredictable_cluster09)
* [UnpredicTable-cluster10](https://huggingface.co/datasets/MicPie/unpredictable_cluster10)
* [UnpredicTable-cluster11](https://huggingface.co/datasets/MicPie/unpredictable_cluster11)
* [UnpredicTable-cluster12](https://huggingface.co/datasets/MicPie/unpredictable_cluster12)
* [UnpredicTable-cluster13](https://huggingface.co/datasets/MicPie/unpredictable_cluster13)
* [UnpredicTable-cluster14](https://huggingface.co/datasets/MicPie/unpredictable_cluster14)
* [UnpredicTable-cluster15](https://huggingface.co/datasets/MicPie/unpredictable_cluster15)
* [UnpredicTable-cluster16](https://huggingface.co/datasets/MicPie/unpredictable_cluster16)
* [UnpredicTable-cluster17](https://huggingface.co/datasets/MicPie/unpredictable_cluster17)
* [UnpredicTable-cluster18](https://huggingface.co/datasets/MicPie/unpredictable_cluster18)
* [UnpredicTable-cluster19](https://huggingface.co/datasets/MicPie/unpredictable_cluster19)
* [UnpredicTable-cluster20](https://huggingface.co/datasets/MicPie/unpredictable_cluster20)
* [UnpredicTable-cluster21](https://huggingface.co/datasets/MicPie/unpredictable_cluster21)
* [UnpredicTable-cluster22](https://huggingface.co/datasets/MicPie/unpredictable_cluster22)
* [UnpredicTable-cluster23](https://huggingface.co/datasets/MicPie/unpredictable_cluster23)
* [UnpredicTable-cluster24](https://huggingface.co/datasets/MicPie/unpredictable_cluster24)
* [UnpredicTable-cluster25](https://huggingface.co/datasets/MicPie/unpredictable_cluster25)
* [UnpredicTable-cluster26](https://huggingface.co/datasets/MicPie/unpredictable_cluster26)
* [UnpredicTable-cluster27](https://huggingface.co/datasets/MicPie/unpredictable_cluster27)
* [UnpredicTable-cluster28](https://huggingface.co/datasets/MicPie/unpredictable_cluster28)
* [UnpredicTable-cluster29](https://huggingface.co/datasets/MicPie/unpredictable_cluster29)
* [UnpredicTable-cluster-noise](https://huggingface.co/datasets/MicPie/unpredictable_cluster-noise)
### Supported Tasks and Leaderboards
Since the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.
The intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.
### Languages
English
## Dataset Structure
### Data Instances
Each task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.
There are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.
### Data Fields
'task': task identifier
'input': column elements of a specific row in the table.
'options': for multiple choice classification, it provides the options to choose from.
'output': target column element of the same row as input.
'pageTitle': the title of the page containing the table.
'outputColName': output column name
'url': url to the website containing the table
'wdcFile': WDC Web Table Corpus file
### Data Splits
The UnpredicTable datasets do not come with additional data splits.
## Dataset Creation
### Curation Rationale
Few-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.
### Source Data
#### Initial Data Collection and Normalization
We use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (http://webdatacommons.org/webtables/2015/EnglishStatistics.html). The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.
#### Who are the source language producers?
The dataset is extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/).
### Annotations
#### Annotation process
Manual annotation was only carried out for the [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low),
[UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium), and [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high) data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.
#### Who are the annotators?
Annotations were carried out by a lab assistant.
### Personal and Sensitive Information
The data was extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/), which in turn extracted tables from the [Common Crawl](https://commoncrawl.org/). We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.
## Considerations for Using the Data
### Social Impact of Dataset
This dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.
### Discussion of Biases
Since our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.
### Other Known Limitations
No additional known limitations.
## Additional Information
### Dataset Curators
Jun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez
### Licensing Information
Apache 2.0
### Citation Information
```
@misc{chan2022few,
author = {Chan, Jun Shern and Pieler, Michael and Jao, Jonathan and Scheurer, Jérémy and Perez, Ethan},
title = {Few-shot Adaptation Works with UnpredicTable Data},
publisher={arXiv},
year = {2022},
url = {https://arxiv.org/abs/2208.01009}
}
```
|
true |
# Dataset Card for "UnpredicTable-cluster15" - Dataset of Few-shot Tasks from Tables
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-instances)
- [Data Splits](#data-instances)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
## Dataset Description
- **Homepage:** https://ethanperez.net/unpredictable
- **Repository:** https://github.com/JunShern/few-shot-adaptation
- **Paper:** Few-shot Adaptation Works with UnpredicTable Data
- **Point of Contact:** junshern@nyu.edu, perez@nyu.edu
### Dataset Summary
The UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.
There are several dataset versions available:
* [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full): Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full), which comprises 413,299 tasks from 23,744 unique websites.
* [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique): This is the same as [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full) but filtered to have a maximum of one task per website. [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique) contains exactly 23,744 tasks from 23,744 websites.
* [UnpredicTable-5k](https://huggingface.co/datasets/MicPie/unpredictable_5k): This dataset contains 5k random tables from the full dataset.
* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):
* [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low)
* [UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium)
* [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high)
* UnpredicTable data subsets based on the website of origin:
* [UnpredicTable-baseball-fantasysports-yahoo-com](https://huggingface.co/datasets/MicPie/unpredictable_baseball-fantasysports-yahoo-com)
* [UnpredicTable-bulbapedia-bulbagarden-net](https://huggingface.co/datasets/MicPie/unpredictable_bulbapedia-bulbagarden-net)
* [UnpredicTable-cappex-com](https://huggingface.co/datasets/MicPie/unpredictable_cappex-com)
* [UnpredicTable-cram-com](https://huggingface.co/datasets/MicPie/unpredictable_cram-com)
* [UnpredicTable-dividend-com](https://huggingface.co/datasets/MicPie/unpredictable_dividend-com)
* [UnpredicTable-dummies-com](https://huggingface.co/datasets/MicPie/unpredictable_dummies-com)
* [UnpredicTable-en-wikipedia-org](https://huggingface.co/datasets/MicPie/unpredictable_en-wikipedia-org)
* [UnpredicTable-ensembl-org](https://huggingface.co/datasets/MicPie/unpredictable_ensembl-org)
* [UnpredicTable-gamefaqs-com](https://huggingface.co/datasets/MicPie/unpredictable_gamefaqs-com)
* [UnpredicTable-mgoblog-com](https://huggingface.co/datasets/MicPie/unpredictable_mgoblog-com)
* [UnpredicTable-mmo-champion-com](https://huggingface.co/datasets/MicPie/unpredictable_mmo-champion-com)
* [UnpredicTable-msdn-microsoft-com](https://huggingface.co/datasets/MicPie/unpredictable_msdn-microsoft-com)
* [UnpredicTable-phonearena-com](https://huggingface.co/datasets/MicPie/unpredictable_phonearena-com)
* [UnpredicTable-sittercity-com](https://huggingface.co/datasets/MicPie/unpredictable_sittercity-com)
* [UnpredicTable-sporcle-com](https://huggingface.co/datasets/MicPie/unpredictable_sporcle-com)
* [UnpredicTable-studystack-com](https://huggingface.co/datasets/MicPie/unpredictable_studystack-com)
* [UnpredicTable-support-google-com](https://huggingface.co/datasets/MicPie/unpredictable_support-google-com)
* [UnpredicTable-w3-org](https://huggingface.co/datasets/MicPie/unpredictable_w3-org)
* [UnpredicTable-wiki-openmoko-org](https://huggingface.co/datasets/MicPie/unpredictable_wiki-openmoko-org)
* [UnpredicTable-wkdu-org](https://huggingface.co/datasets/MicPie/unpredictable_wkdu-org)
* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):
* [UnpredicTable-cluster00](https://huggingface.co/datasets/MicPie/unpredictable_cluster00)
* [UnpredicTable-cluster01](https://huggingface.co/datasets/MicPie/unpredictable_cluster01)
* [UnpredicTable-cluster02](https://huggingface.co/datasets/MicPie/unpredictable_cluster02)
* [UnpredicTable-cluster03](https://huggingface.co/datasets/MicPie/unpredictable_cluster03)
* [UnpredicTable-cluster04](https://huggingface.co/datasets/MicPie/unpredictable_cluster04)
* [UnpredicTable-cluster05](https://huggingface.co/datasets/MicPie/unpredictable_cluster05)
* [UnpredicTable-cluster06](https://huggingface.co/datasets/MicPie/unpredictable_cluster06)
* [UnpredicTable-cluster07](https://huggingface.co/datasets/MicPie/unpredictable_cluster07)
* [UnpredicTable-cluster08](https://huggingface.co/datasets/MicPie/unpredictable_cluster08)
* [UnpredicTable-cluster09](https://huggingface.co/datasets/MicPie/unpredictable_cluster09)
* [UnpredicTable-cluster10](https://huggingface.co/datasets/MicPie/unpredictable_cluster10)
* [UnpredicTable-cluster11](https://huggingface.co/datasets/MicPie/unpredictable_cluster11)
* [UnpredicTable-cluster12](https://huggingface.co/datasets/MicPie/unpredictable_cluster12)
* [UnpredicTable-cluster13](https://huggingface.co/datasets/MicPie/unpredictable_cluster13)
* [UnpredicTable-cluster14](https://huggingface.co/datasets/MicPie/unpredictable_cluster14)
* [UnpredicTable-cluster15](https://huggingface.co/datasets/MicPie/unpredictable_cluster15)
* [UnpredicTable-cluster16](https://huggingface.co/datasets/MicPie/unpredictable_cluster16)
* [UnpredicTable-cluster17](https://huggingface.co/datasets/MicPie/unpredictable_cluster17)
* [UnpredicTable-cluster18](https://huggingface.co/datasets/MicPie/unpredictable_cluster18)
* [UnpredicTable-cluster19](https://huggingface.co/datasets/MicPie/unpredictable_cluster19)
* [UnpredicTable-cluster20](https://huggingface.co/datasets/MicPie/unpredictable_cluster20)
* [UnpredicTable-cluster21](https://huggingface.co/datasets/MicPie/unpredictable_cluster21)
* [UnpredicTable-cluster22](https://huggingface.co/datasets/MicPie/unpredictable_cluster22)
* [UnpredicTable-cluster23](https://huggingface.co/datasets/MicPie/unpredictable_cluster23)
* [UnpredicTable-cluster24](https://huggingface.co/datasets/MicPie/unpredictable_cluster24)
* [UnpredicTable-cluster25](https://huggingface.co/datasets/MicPie/unpredictable_cluster25)
* [UnpredicTable-cluster26](https://huggingface.co/datasets/MicPie/unpredictable_cluster26)
* [UnpredicTable-cluster27](https://huggingface.co/datasets/MicPie/unpredictable_cluster27)
* [UnpredicTable-cluster28](https://huggingface.co/datasets/MicPie/unpredictable_cluster28)
* [UnpredicTable-cluster29](https://huggingface.co/datasets/MicPie/unpredictable_cluster29)
* [UnpredicTable-cluster-noise](https://huggingface.co/datasets/MicPie/unpredictable_cluster-noise)
### Supported Tasks and Leaderboards
Since the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.
The intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.
### Languages
English
## Dataset Structure
### Data Instances
Each task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.
There are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.
### Data Fields
'task': task identifier
'input': column elements of a specific row in the table.
'options': for multiple choice classification, it provides the options to choose from.
'output': target column element of the same row as input.
'pageTitle': the title of the page containing the table.
'outputColName': output column name
'url': url to the website containing the table
'wdcFile': WDC Web Table Corpus file
### Data Splits
The UnpredicTable datasets do not come with additional data splits.
## Dataset Creation
### Curation Rationale
Few-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.
### Source Data
#### Initial Data Collection and Normalization
We use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (http://webdatacommons.org/webtables/2015/EnglishStatistics.html). The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.
#### Who are the source language producers?
The dataset is extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/).
### Annotations
#### Annotation process
Manual annotation was only carried out for the [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low),
[UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium), and [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high) data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.
#### Who are the annotators?
Annotations were carried out by a lab assistant.
### Personal and Sensitive Information
The data was extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/), which in turn extracted tables from the [Common Crawl](https://commoncrawl.org/). We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.
## Considerations for Using the Data
### Social Impact of Dataset
This dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.
### Discussion of Biases
Since our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.
### Other Known Limitations
No additional known limitations.
## Additional Information
### Dataset Curators
Jun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez
### Licensing Information
Apache 2.0
### Citation Information
```
@misc{chan2022few,
author = {Chan, Jun Shern and Pieler, Michael and Jao, Jonathan and Scheurer, Jérémy and Perez, Ethan},
title = {Few-shot Adaptation Works with UnpredicTable Data},
publisher={arXiv},
year = {2022},
url = {https://arxiv.org/abs/2208.01009}
}
```
|
true |
# Dataset Card for "UnpredicTable-cluster17" - Dataset of Few-shot Tasks from Tables
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-instances)
- [Data Splits](#data-instances)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
## Dataset Description
- **Homepage:** https://ethanperez.net/unpredictable
- **Repository:** https://github.com/JunShern/few-shot-adaptation
- **Paper:** Few-shot Adaptation Works with UnpredicTable Data
- **Point of Contact:** junshern@nyu.edu, perez@nyu.edu
### Dataset Summary
The UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.
There are several dataset versions available:
* [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full): Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full), which comprises 413,299 tasks from 23,744 unique websites.
* [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique): This is the same as [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full) but filtered to have a maximum of one task per website. [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique) contains exactly 23,744 tasks from 23,744 websites.
* [UnpredicTable-5k](https://huggingface.co/datasets/MicPie/unpredictable_5k): This dataset contains 5k random tables from the full dataset.
* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):
* [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low)
* [UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium)
* [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high)
* UnpredicTable data subsets based on the website of origin:
* [UnpredicTable-baseball-fantasysports-yahoo-com](https://huggingface.co/datasets/MicPie/unpredictable_baseball-fantasysports-yahoo-com)
* [UnpredicTable-bulbapedia-bulbagarden-net](https://huggingface.co/datasets/MicPie/unpredictable_bulbapedia-bulbagarden-net)
* [UnpredicTable-cappex-com](https://huggingface.co/datasets/MicPie/unpredictable_cappex-com)
* [UnpredicTable-cram-com](https://huggingface.co/datasets/MicPie/unpredictable_cram-com)
* [UnpredicTable-dividend-com](https://huggingface.co/datasets/MicPie/unpredictable_dividend-com)
* [UnpredicTable-dummies-com](https://huggingface.co/datasets/MicPie/unpredictable_dummies-com)
* [UnpredicTable-en-wikipedia-org](https://huggingface.co/datasets/MicPie/unpredictable_en-wikipedia-org)
* [UnpredicTable-ensembl-org](https://huggingface.co/datasets/MicPie/unpredictable_ensembl-org)
* [UnpredicTable-gamefaqs-com](https://huggingface.co/datasets/MicPie/unpredictable_gamefaqs-com)
* [UnpredicTable-mgoblog-com](https://huggingface.co/datasets/MicPie/unpredictable_mgoblog-com)
* [UnpredicTable-mmo-champion-com](https://huggingface.co/datasets/MicPie/unpredictable_mmo-champion-com)
* [UnpredicTable-msdn-microsoft-com](https://huggingface.co/datasets/MicPie/unpredictable_msdn-microsoft-com)
* [UnpredicTable-phonearena-com](https://huggingface.co/datasets/MicPie/unpredictable_phonearena-com)
* [UnpredicTable-sittercity-com](https://huggingface.co/datasets/MicPie/unpredictable_sittercity-com)
* [UnpredicTable-sporcle-com](https://huggingface.co/datasets/MicPie/unpredictable_sporcle-com)
* [UnpredicTable-studystack-com](https://huggingface.co/datasets/MicPie/unpredictable_studystack-com)
* [UnpredicTable-support-google-com](https://huggingface.co/datasets/MicPie/unpredictable_support-google-com)
* [UnpredicTable-w3-org](https://huggingface.co/datasets/MicPie/unpredictable_w3-org)
* [UnpredicTable-wiki-openmoko-org](https://huggingface.co/datasets/MicPie/unpredictable_wiki-openmoko-org)
* [UnpredicTable-wkdu-org](https://huggingface.co/datasets/MicPie/unpredictable_wkdu-org)
* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):
* [UnpredicTable-cluster00](https://huggingface.co/datasets/MicPie/unpredictable_cluster00)
* [UnpredicTable-cluster01](https://huggingface.co/datasets/MicPie/unpredictable_cluster01)
* [UnpredicTable-cluster02](https://huggingface.co/datasets/MicPie/unpredictable_cluster02)
* [UnpredicTable-cluster03](https://huggingface.co/datasets/MicPie/unpredictable_cluster03)
* [UnpredicTable-cluster04](https://huggingface.co/datasets/MicPie/unpredictable_cluster04)
* [UnpredicTable-cluster05](https://huggingface.co/datasets/MicPie/unpredictable_cluster05)
* [UnpredicTable-cluster06](https://huggingface.co/datasets/MicPie/unpredictable_cluster06)
* [UnpredicTable-cluster07](https://huggingface.co/datasets/MicPie/unpredictable_cluster07)
* [UnpredicTable-cluster08](https://huggingface.co/datasets/MicPie/unpredictable_cluster08)
* [UnpredicTable-cluster09](https://huggingface.co/datasets/MicPie/unpredictable_cluster09)
* [UnpredicTable-cluster10](https://huggingface.co/datasets/MicPie/unpredictable_cluster10)
* [UnpredicTable-cluster11](https://huggingface.co/datasets/MicPie/unpredictable_cluster11)
* [UnpredicTable-cluster12](https://huggingface.co/datasets/MicPie/unpredictable_cluster12)
* [UnpredicTable-cluster13](https://huggingface.co/datasets/MicPie/unpredictable_cluster13)
* [UnpredicTable-cluster14](https://huggingface.co/datasets/MicPie/unpredictable_cluster14)
* [UnpredicTable-cluster15](https://huggingface.co/datasets/MicPie/unpredictable_cluster15)
* [UnpredicTable-cluster16](https://huggingface.co/datasets/MicPie/unpredictable_cluster16)
* [UnpredicTable-cluster17](https://huggingface.co/datasets/MicPie/unpredictable_cluster17)
* [UnpredicTable-cluster18](https://huggingface.co/datasets/MicPie/unpredictable_cluster18)
* [UnpredicTable-cluster19](https://huggingface.co/datasets/MicPie/unpredictable_cluster19)
* [UnpredicTable-cluster20](https://huggingface.co/datasets/MicPie/unpredictable_cluster20)
* [UnpredicTable-cluster21](https://huggingface.co/datasets/MicPie/unpredictable_cluster21)
* [UnpredicTable-cluster22](https://huggingface.co/datasets/MicPie/unpredictable_cluster22)
* [UnpredicTable-cluster23](https://huggingface.co/datasets/MicPie/unpredictable_cluster23)
* [UnpredicTable-cluster24](https://huggingface.co/datasets/MicPie/unpredictable_cluster24)
* [UnpredicTable-cluster25](https://huggingface.co/datasets/MicPie/unpredictable_cluster25)
* [UnpredicTable-cluster26](https://huggingface.co/datasets/MicPie/unpredictable_cluster26)
* [UnpredicTable-cluster27](https://huggingface.co/datasets/MicPie/unpredictable_cluster27)
* [UnpredicTable-cluster28](https://huggingface.co/datasets/MicPie/unpredictable_cluster28)
* [UnpredicTable-cluster29](https://huggingface.co/datasets/MicPie/unpredictable_cluster29)
* [UnpredicTable-cluster-noise](https://huggingface.co/datasets/MicPie/unpredictable_cluster-noise)
### Supported Tasks and Leaderboards
Since the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.
The intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.
### Languages
English
## Dataset Structure
### Data Instances
Each task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.
There are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.
### Data Fields
'task': task identifier
'input': column elements of a specific row in the table.
'options': for multiple choice classification, it provides the options to choose from.
'output': target column element of the same row as input.
'pageTitle': the title of the page containing the table.
'outputColName': output column name
'url': url to the website containing the table
'wdcFile': WDC Web Table Corpus file
### Data Splits
The UnpredicTable datasets do not come with additional data splits.
## Dataset Creation
### Curation Rationale
Few-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.
### Source Data
#### Initial Data Collection and Normalization
We use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (http://webdatacommons.org/webtables/2015/EnglishStatistics.html). The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.
#### Who are the source language producers?
The dataset is extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/).
### Annotations
#### Annotation process
Manual annotation was only carried out for the [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low),
[UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium), and [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high) data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.
#### Who are the annotators?
Annotations were carried out by a lab assistant.
### Personal and Sensitive Information
The data was extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/), which in turn extracted tables from the [Common Crawl](https://commoncrawl.org/). We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.
## Considerations for Using the Data
### Social Impact of Dataset
This dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.
### Discussion of Biases
Since our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.
### Other Known Limitations
No additional known limitations.
## Additional Information
### Dataset Curators
Jun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez
### Licensing Information
Apache 2.0
### Citation Information
```
@misc{chan2022few,
author = {Chan, Jun Shern and Pieler, Michael and Jao, Jonathan and Scheurer, Jérémy and Perez, Ethan},
title = {Few-shot Adaptation Works with UnpredicTable Data},
publisher={arXiv},
year = {2022},
url = {https://arxiv.org/abs/2208.01009}
}
```
|
true |
# Dataset Card for "UnpredicTable-cluster24" - Dataset of Few-shot Tasks from Tables
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-instances)
- [Data Splits](#data-instances)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
## Dataset Description
- **Homepage:** https://ethanperez.net/unpredictable
- **Repository:** https://github.com/JunShern/few-shot-adaptation
- **Paper:** Few-shot Adaptation Works with UnpredicTable Data
- **Point of Contact:** junshern@nyu.edu, perez@nyu.edu
### Dataset Summary
The UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.
There are several dataset versions available:
* [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full): Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full), which comprises 413,299 tasks from 23,744 unique websites.
* [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique): This is the same as [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full) but filtered to have a maximum of one task per website. [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique) contains exactly 23,744 tasks from 23,744 websites.
* [UnpredicTable-5k](https://huggingface.co/datasets/MicPie/unpredictable_5k): This dataset contains 5k random tables from the full dataset.
* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):
* [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low)
* [UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium)
* [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high)
* UnpredicTable data subsets based on the website of origin:
* [UnpredicTable-baseball-fantasysports-yahoo-com](https://huggingface.co/datasets/MicPie/unpredictable_baseball-fantasysports-yahoo-com)
* [UnpredicTable-bulbapedia-bulbagarden-net](https://huggingface.co/datasets/MicPie/unpredictable_bulbapedia-bulbagarden-net)
* [UnpredicTable-cappex-com](https://huggingface.co/datasets/MicPie/unpredictable_cappex-com)
* [UnpredicTable-cram-com](https://huggingface.co/datasets/MicPie/unpredictable_cram-com)
* [UnpredicTable-dividend-com](https://huggingface.co/datasets/MicPie/unpredictable_dividend-com)
* [UnpredicTable-dummies-com](https://huggingface.co/datasets/MicPie/unpredictable_dummies-com)
* [UnpredicTable-en-wikipedia-org](https://huggingface.co/datasets/MicPie/unpredictable_en-wikipedia-org)
* [UnpredicTable-ensembl-org](https://huggingface.co/datasets/MicPie/unpredictable_ensembl-org)
* [UnpredicTable-gamefaqs-com](https://huggingface.co/datasets/MicPie/unpredictable_gamefaqs-com)
* [UnpredicTable-mgoblog-com](https://huggingface.co/datasets/MicPie/unpredictable_mgoblog-com)
* [UnpredicTable-mmo-champion-com](https://huggingface.co/datasets/MicPie/unpredictable_mmo-champion-com)
* [UnpredicTable-msdn-microsoft-com](https://huggingface.co/datasets/MicPie/unpredictable_msdn-microsoft-com)
* [UnpredicTable-phonearena-com](https://huggingface.co/datasets/MicPie/unpredictable_phonearena-com)
* [UnpredicTable-sittercity-com](https://huggingface.co/datasets/MicPie/unpredictable_sittercity-com)
* [UnpredicTable-sporcle-com](https://huggingface.co/datasets/MicPie/unpredictable_sporcle-com)
* [UnpredicTable-studystack-com](https://huggingface.co/datasets/MicPie/unpredictable_studystack-com)
* [UnpredicTable-support-google-com](https://huggingface.co/datasets/MicPie/unpredictable_support-google-com)
* [UnpredicTable-w3-org](https://huggingface.co/datasets/MicPie/unpredictable_w3-org)
* [UnpredicTable-wiki-openmoko-org](https://huggingface.co/datasets/MicPie/unpredictable_wiki-openmoko-org)
* [UnpredicTable-wkdu-org](https://huggingface.co/datasets/MicPie/unpredictable_wkdu-org)
* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):
* [UnpredicTable-cluster00](https://huggingface.co/datasets/MicPie/unpredictable_cluster00)
* [UnpredicTable-cluster01](https://huggingface.co/datasets/MicPie/unpredictable_cluster01)
* [UnpredicTable-cluster02](https://huggingface.co/datasets/MicPie/unpredictable_cluster02)
* [UnpredicTable-cluster03](https://huggingface.co/datasets/MicPie/unpredictable_cluster03)
* [UnpredicTable-cluster04](https://huggingface.co/datasets/MicPie/unpredictable_cluster04)
* [UnpredicTable-cluster05](https://huggingface.co/datasets/MicPie/unpredictable_cluster05)
* [UnpredicTable-cluster06](https://huggingface.co/datasets/MicPie/unpredictable_cluster06)
* [UnpredicTable-cluster07](https://huggingface.co/datasets/MicPie/unpredictable_cluster07)
* [UnpredicTable-cluster08](https://huggingface.co/datasets/MicPie/unpredictable_cluster08)
* [UnpredicTable-cluster09](https://huggingface.co/datasets/MicPie/unpredictable_cluster09)
* [UnpredicTable-cluster10](https://huggingface.co/datasets/MicPie/unpredictable_cluster10)
* [UnpredicTable-cluster11](https://huggingface.co/datasets/MicPie/unpredictable_cluster11)
* [UnpredicTable-cluster12](https://huggingface.co/datasets/MicPie/unpredictable_cluster12)
* [UnpredicTable-cluster13](https://huggingface.co/datasets/MicPie/unpredictable_cluster13)
* [UnpredicTable-cluster14](https://huggingface.co/datasets/MicPie/unpredictable_cluster14)
* [UnpredicTable-cluster15](https://huggingface.co/datasets/MicPie/unpredictable_cluster15)
* [UnpredicTable-cluster16](https://huggingface.co/datasets/MicPie/unpredictable_cluster16)
* [UnpredicTable-cluster17](https://huggingface.co/datasets/MicPie/unpredictable_cluster17)
* [UnpredicTable-cluster18](https://huggingface.co/datasets/MicPie/unpredictable_cluster18)
* [UnpredicTable-cluster19](https://huggingface.co/datasets/MicPie/unpredictable_cluster19)
* [UnpredicTable-cluster20](https://huggingface.co/datasets/MicPie/unpredictable_cluster20)
* [UnpredicTable-cluster21](https://huggingface.co/datasets/MicPie/unpredictable_cluster21)
* [UnpredicTable-cluster22](https://huggingface.co/datasets/MicPie/unpredictable_cluster22)
* [UnpredicTable-cluster23](https://huggingface.co/datasets/MicPie/unpredictable_cluster23)
* [UnpredicTable-cluster24](https://huggingface.co/datasets/MicPie/unpredictable_cluster24)
* [UnpredicTable-cluster25](https://huggingface.co/datasets/MicPie/unpredictable_cluster25)
* [UnpredicTable-cluster26](https://huggingface.co/datasets/MicPie/unpredictable_cluster26)
* [UnpredicTable-cluster27](https://huggingface.co/datasets/MicPie/unpredictable_cluster27)
* [UnpredicTable-cluster28](https://huggingface.co/datasets/MicPie/unpredictable_cluster28)
* [UnpredicTable-cluster29](https://huggingface.co/datasets/MicPie/unpredictable_cluster29)
* [UnpredicTable-cluster-noise](https://huggingface.co/datasets/MicPie/unpredictable_cluster-noise)
### Supported Tasks and Leaderboards
Since the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.
The intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.
### Languages
English
## Dataset Structure
### Data Instances
Each task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.
There are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.
### Data Fields
'task': task identifier
'input': column elements of a specific row in the table.
'options': for multiple choice classification, it provides the options to choose from.
'output': target column element of the same row as input.
'pageTitle': the title of the page containing the table.
'outputColName': output column name
'url': url to the website containing the table
'wdcFile': WDC Web Table Corpus file
### Data Splits
The UnpredicTable datasets do not come with additional data splits.
## Dataset Creation
### Curation Rationale
Few-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.
### Source Data
#### Initial Data Collection and Normalization
We use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (http://webdatacommons.org/webtables/2015/EnglishStatistics.html). The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.
#### Who are the source language producers?
The dataset is extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/).
### Annotations
#### Annotation process
Manual annotation was only carried out for the [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low),
[UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium), and [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high) data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.
#### Who are the annotators?
Annotations were carried out by a lab assistant.
### Personal and Sensitive Information
The data was extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/), which in turn extracted tables from the [Common Crawl](https://commoncrawl.org/). We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.
## Considerations for Using the Data
### Social Impact of Dataset
This dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.
### Discussion of Biases
Since our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.
### Other Known Limitations
No additional known limitations.
## Additional Information
### Dataset Curators
Jun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez
### Licensing Information
Apache 2.0
### Citation Information
```
@misc{chan2022few,
author = {Chan, Jun Shern and Pieler, Michael and Jao, Jonathan and Scheurer, Jérémy and Perez, Ethan},
title = {Few-shot Adaptation Works with UnpredicTable Data},
publisher={arXiv},
year = {2022},
url = {https://arxiv.org/abs/2208.01009}
}
```
|
true |
# Dataset Card for "UnpredicTable-cluster29" - Dataset of Few-shot Tasks from Tables
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-instances)
- [Data Splits](#data-instances)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
## Dataset Description
- **Homepage:** https://ethanperez.net/unpredictable
- **Repository:** https://github.com/JunShern/few-shot-adaptation
- **Paper:** Few-shot Adaptation Works with UnpredicTable Data
- **Point of Contact:** junshern@nyu.edu, perez@nyu.edu
### Dataset Summary
The UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.
There are several dataset versions available:
* [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full): Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full), which comprises 413,299 tasks from 23,744 unique websites.
* [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique): This is the same as [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full) but filtered to have a maximum of one task per website. [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique) contains exactly 23,744 tasks from 23,744 websites.
* [UnpredicTable-5k](https://huggingface.co/datasets/MicPie/unpredictable_5k): This dataset contains 5k random tables from the full dataset.
* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):
* [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low)
* [UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium)
* [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high)
* UnpredicTable data subsets based on the website of origin:
* [UnpredicTable-baseball-fantasysports-yahoo-com](https://huggingface.co/datasets/MicPie/unpredictable_baseball-fantasysports-yahoo-com)
* [UnpredicTable-bulbapedia-bulbagarden-net](https://huggingface.co/datasets/MicPie/unpredictable_bulbapedia-bulbagarden-net)
* [UnpredicTable-cappex-com](https://huggingface.co/datasets/MicPie/unpredictable_cappex-com)
* [UnpredicTable-cram-com](https://huggingface.co/datasets/MicPie/unpredictable_cram-com)
* [UnpredicTable-dividend-com](https://huggingface.co/datasets/MicPie/unpredictable_dividend-com)
* [UnpredicTable-dummies-com](https://huggingface.co/datasets/MicPie/unpredictable_dummies-com)
* [UnpredicTable-en-wikipedia-org](https://huggingface.co/datasets/MicPie/unpredictable_en-wikipedia-org)
* [UnpredicTable-ensembl-org](https://huggingface.co/datasets/MicPie/unpredictable_ensembl-org)
* [UnpredicTable-gamefaqs-com](https://huggingface.co/datasets/MicPie/unpredictable_gamefaqs-com)
* [UnpredicTable-mgoblog-com](https://huggingface.co/datasets/MicPie/unpredictable_mgoblog-com)
* [UnpredicTable-mmo-champion-com](https://huggingface.co/datasets/MicPie/unpredictable_mmo-champion-com)
* [UnpredicTable-msdn-microsoft-com](https://huggingface.co/datasets/MicPie/unpredictable_msdn-microsoft-com)
* [UnpredicTable-phonearena-com](https://huggingface.co/datasets/MicPie/unpredictable_phonearena-com)
* [UnpredicTable-sittercity-com](https://huggingface.co/datasets/MicPie/unpredictable_sittercity-com)
* [UnpredicTable-sporcle-com](https://huggingface.co/datasets/MicPie/unpredictable_sporcle-com)
* [UnpredicTable-studystack-com](https://huggingface.co/datasets/MicPie/unpredictable_studystack-com)
* [UnpredicTable-support-google-com](https://huggingface.co/datasets/MicPie/unpredictable_support-google-com)
* [UnpredicTable-w3-org](https://huggingface.co/datasets/MicPie/unpredictable_w3-org)
* [UnpredicTable-wiki-openmoko-org](https://huggingface.co/datasets/MicPie/unpredictable_wiki-openmoko-org)
* [UnpredicTable-wkdu-org](https://huggingface.co/datasets/MicPie/unpredictable_wkdu-org)
* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):
* [UnpredicTable-cluster00](https://huggingface.co/datasets/MicPie/unpredictable_cluster00)
* [UnpredicTable-cluster01](https://huggingface.co/datasets/MicPie/unpredictable_cluster01)
* [UnpredicTable-cluster02](https://huggingface.co/datasets/MicPie/unpredictable_cluster02)
* [UnpredicTable-cluster03](https://huggingface.co/datasets/MicPie/unpredictable_cluster03)
* [UnpredicTable-cluster04](https://huggingface.co/datasets/MicPie/unpredictable_cluster04)
* [UnpredicTable-cluster05](https://huggingface.co/datasets/MicPie/unpredictable_cluster05)
* [UnpredicTable-cluster06](https://huggingface.co/datasets/MicPie/unpredictable_cluster06)
* [UnpredicTable-cluster07](https://huggingface.co/datasets/MicPie/unpredictable_cluster07)
* [UnpredicTable-cluster08](https://huggingface.co/datasets/MicPie/unpredictable_cluster08)
* [UnpredicTable-cluster09](https://huggingface.co/datasets/MicPie/unpredictable_cluster09)
* [UnpredicTable-cluster10](https://huggingface.co/datasets/MicPie/unpredictable_cluster10)
* [UnpredicTable-cluster11](https://huggingface.co/datasets/MicPie/unpredictable_cluster11)
* [UnpredicTable-cluster12](https://huggingface.co/datasets/MicPie/unpredictable_cluster12)
* [UnpredicTable-cluster13](https://huggingface.co/datasets/MicPie/unpredictable_cluster13)
* [UnpredicTable-cluster14](https://huggingface.co/datasets/MicPie/unpredictable_cluster14)
* [UnpredicTable-cluster15](https://huggingface.co/datasets/MicPie/unpredictable_cluster15)
* [UnpredicTable-cluster16](https://huggingface.co/datasets/MicPie/unpredictable_cluster16)
* [UnpredicTable-cluster17](https://huggingface.co/datasets/MicPie/unpredictable_cluster17)
* [UnpredicTable-cluster18](https://huggingface.co/datasets/MicPie/unpredictable_cluster18)
* [UnpredicTable-cluster19](https://huggingface.co/datasets/MicPie/unpredictable_cluster19)
* [UnpredicTable-cluster20](https://huggingface.co/datasets/MicPie/unpredictable_cluster20)
* [UnpredicTable-cluster21](https://huggingface.co/datasets/MicPie/unpredictable_cluster21)
* [UnpredicTable-cluster22](https://huggingface.co/datasets/MicPie/unpredictable_cluster22)
* [UnpredicTable-cluster23](https://huggingface.co/datasets/MicPie/unpredictable_cluster23)
* [UnpredicTable-cluster24](https://huggingface.co/datasets/MicPie/unpredictable_cluster24)
* [UnpredicTable-cluster25](https://huggingface.co/datasets/MicPie/unpredictable_cluster25)
* [UnpredicTable-cluster26](https://huggingface.co/datasets/MicPie/unpredictable_cluster26)
* [UnpredicTable-cluster27](https://huggingface.co/datasets/MicPie/unpredictable_cluster27)
* [UnpredicTable-cluster28](https://huggingface.co/datasets/MicPie/unpredictable_cluster28)
* [UnpredicTable-cluster29](https://huggingface.co/datasets/MicPie/unpredictable_cluster29)
* [UnpredicTable-cluster-noise](https://huggingface.co/datasets/MicPie/unpredictable_cluster-noise)
### Supported Tasks and Leaderboards
Since the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.
The intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.
### Languages
English
## Dataset Structure
### Data Instances
Each task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.
There are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.
### Data Fields
'task': task identifier
'input': column elements of a specific row in the table.
'options': for multiple choice classification, it provides the options to choose from.
'output': target column element of the same row as input.
'pageTitle': the title of the page containing the table.
'outputColName': output column name
'url': url to the website containing the table
'wdcFile': WDC Web Table Corpus file
### Data Splits
The UnpredicTable datasets do not come with additional data splits.
## Dataset Creation
### Curation Rationale
Few-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.
### Source Data
#### Initial Data Collection and Normalization
We use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (http://webdatacommons.org/webtables/2015/EnglishStatistics.html). The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.
#### Who are the source language producers?
The dataset is extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/).
### Annotations
#### Annotation process
Manual annotation was only carried out for the [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low),
[UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium), and [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high) data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.
#### Who are the annotators?
Annotations were carried out by a lab assistant.
### Personal and Sensitive Information
The data was extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/), which in turn extracted tables from the [Common Crawl](https://commoncrawl.org/). We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.
## Considerations for Using the Data
### Social Impact of Dataset
This dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.
### Discussion of Biases
Since our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.
### Other Known Limitations
No additional known limitations.
## Additional Information
### Dataset Curators
Jun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez
### Licensing Information
Apache 2.0
### Citation Information
```
@misc{chan2022few,
author = {Chan, Jun Shern and Pieler, Michael and Jao, Jonathan and Scheurer, Jérémy and Perez, Ethan},
title = {Few-shot Adaptation Works with UnpredicTable Data},
publisher={arXiv},
year = {2022},
url = {https://arxiv.org/abs/2208.01009}
}
```
|
true |
# Dataset Card for "UnpredicTable-cluster06" - Dataset of Few-shot Tasks from Tables
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-instances)
- [Data Splits](#data-instances)
- [Dataset Creation](#dataset-creation)
- [Curation Rationale](#curation-rationale)
- [Source Data](#source-data)
- [Annotations](#annotations)
- [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Social Impact of Dataset](#social-impact-of-dataset)
- [Discussion of Biases](#discussion-of-biases)
- [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
- [Dataset Curators](#dataset-curators)
- [Licensing Information](#licensing-information)
- [Citation Information](#citation-information)
## Dataset Description
- **Homepage:** https://ethanperez.net/unpredictable
- **Repository:** https://github.com/JunShern/few-shot-adaptation
- **Paper:** Few-shot Adaptation Works with UnpredicTable Data
- **Point of Contact:** junshern@nyu.edu, perez@nyu.edu
### Dataset Summary
The UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.
There are several dataset versions available:
* [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full): Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full), which comprises 413,299 tasks from 23,744 unique websites.
* [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique): This is the same as [UnpredicTable-full](https://huggingface.co/datasets/MicPie/unpredictable_full) but filtered to have a maximum of one task per website. [UnpredicTable-unique](https://huggingface.co/datasets/MicPie/unpredictable_unique) contains exactly 23,744 tasks from 23,744 websites.
* [UnpredicTable-5k](https://huggingface.co/datasets/MicPie/unpredictable_5k): This dataset contains 5k random tables from the full dataset.
* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):
* [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low)
* [UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium)
* [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high)
* UnpredicTable data subsets based on the website of origin:
* [UnpredicTable-baseball-fantasysports-yahoo-com](https://huggingface.co/datasets/MicPie/unpredictable_baseball-fantasysports-yahoo-com)
* [UnpredicTable-bulbapedia-bulbagarden-net](https://huggingface.co/datasets/MicPie/unpredictable_bulbapedia-bulbagarden-net)
* [UnpredicTable-cappex-com](https://huggingface.co/datasets/MicPie/unpredictable_cappex-com)
* [UnpredicTable-cram-com](https://huggingface.co/datasets/MicPie/unpredictable_cram-com)
* [UnpredicTable-dividend-com](https://huggingface.co/datasets/MicPie/unpredictable_dividend-com)
* [UnpredicTable-dummies-com](https://huggingface.co/datasets/MicPie/unpredictable_dummies-com)
* [UnpredicTable-en-wikipedia-org](https://huggingface.co/datasets/MicPie/unpredictable_en-wikipedia-org)
* [UnpredicTable-ensembl-org](https://huggingface.co/datasets/MicPie/unpredictable_ensembl-org)
* [UnpredicTable-gamefaqs-com](https://huggingface.co/datasets/MicPie/unpredictable_gamefaqs-com)
* [UnpredicTable-mgoblog-com](https://huggingface.co/datasets/MicPie/unpredictable_mgoblog-com)
* [UnpredicTable-mmo-champion-com](https://huggingface.co/datasets/MicPie/unpredictable_mmo-champion-com)
* [UnpredicTable-msdn-microsoft-com](https://huggingface.co/datasets/MicPie/unpredictable_msdn-microsoft-com)
* [UnpredicTable-phonearena-com](https://huggingface.co/datasets/MicPie/unpredictable_phonearena-com)
* [UnpredicTable-sittercity-com](https://huggingface.co/datasets/MicPie/unpredictable_sittercity-com)
* [UnpredicTable-sporcle-com](https://huggingface.co/datasets/MicPie/unpredictable_sporcle-com)
* [UnpredicTable-studystack-com](https://huggingface.co/datasets/MicPie/unpredictable_studystack-com)
* [UnpredicTable-support-google-com](https://huggingface.co/datasets/MicPie/unpredictable_support-google-com)
* [UnpredicTable-w3-org](https://huggingface.co/datasets/MicPie/unpredictable_w3-org)
* [UnpredicTable-wiki-openmoko-org](https://huggingface.co/datasets/MicPie/unpredictable_wiki-openmoko-org)
* [UnpredicTable-wkdu-org](https://huggingface.co/datasets/MicPie/unpredictable_wkdu-org)
* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):
* [UnpredicTable-cluster00](https://huggingface.co/datasets/MicPie/unpredictable_cluster00)
* [UnpredicTable-cluster01](https://huggingface.co/datasets/MicPie/unpredictable_cluster01)
* [UnpredicTable-cluster02](https://huggingface.co/datasets/MicPie/unpredictable_cluster02)
* [UnpredicTable-cluster03](https://huggingface.co/datasets/MicPie/unpredictable_cluster03)
* [UnpredicTable-cluster04](https://huggingface.co/datasets/MicPie/unpredictable_cluster04)
* [UnpredicTable-cluster05](https://huggingface.co/datasets/MicPie/unpredictable_cluster05)
* [UnpredicTable-cluster06](https://huggingface.co/datasets/MicPie/unpredictable_cluster06)
* [UnpredicTable-cluster07](https://huggingface.co/datasets/MicPie/unpredictable_cluster07)
* [UnpredicTable-cluster08](https://huggingface.co/datasets/MicPie/unpredictable_cluster08)
* [UnpredicTable-cluster09](https://huggingface.co/datasets/MicPie/unpredictable_cluster09)
* [UnpredicTable-cluster10](https://huggingface.co/datasets/MicPie/unpredictable_cluster10)
* [UnpredicTable-cluster11](https://huggingface.co/datasets/MicPie/unpredictable_cluster11)
* [UnpredicTable-cluster12](https://huggingface.co/datasets/MicPie/unpredictable_cluster12)
* [UnpredicTable-cluster13](https://huggingface.co/datasets/MicPie/unpredictable_cluster13)
* [UnpredicTable-cluster14](https://huggingface.co/datasets/MicPie/unpredictable_cluster14)
* [UnpredicTable-cluster15](https://huggingface.co/datasets/MicPie/unpredictable_cluster15)
* [UnpredicTable-cluster16](https://huggingface.co/datasets/MicPie/unpredictable_cluster16)
* [UnpredicTable-cluster17](https://huggingface.co/datasets/MicPie/unpredictable_cluster17)
* [UnpredicTable-cluster18](https://huggingface.co/datasets/MicPie/unpredictable_cluster18)
* [UnpredicTable-cluster19](https://huggingface.co/datasets/MicPie/unpredictable_cluster19)
* [UnpredicTable-cluster20](https://huggingface.co/datasets/MicPie/unpredictable_cluster20)
* [UnpredicTable-cluster21](https://huggingface.co/datasets/MicPie/unpredictable_cluster21)
* [UnpredicTable-cluster22](https://huggingface.co/datasets/MicPie/unpredictable_cluster22)
* [UnpredicTable-cluster23](https://huggingface.co/datasets/MicPie/unpredictable_cluster23)
* [UnpredicTable-cluster24](https://huggingface.co/datasets/MicPie/unpredictable_cluster24)
* [UnpredicTable-cluster25](https://huggingface.co/datasets/MicPie/unpredictable_cluster25)
* [UnpredicTable-cluster26](https://huggingface.co/datasets/MicPie/unpredictable_cluster26)
* [UnpredicTable-cluster27](https://huggingface.co/datasets/MicPie/unpredictable_cluster27)
* [UnpredicTable-cluster28](https://huggingface.co/datasets/MicPie/unpredictable_cluster28)
* [UnpredicTable-cluster29](https://huggingface.co/datasets/MicPie/unpredictable_cluster29)
* [UnpredicTable-cluster-noise](https://huggingface.co/datasets/MicPie/unpredictable_cluster-noise)
### Supported Tasks and Leaderboards
Since the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.
The intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.
### Languages
English
## Dataset Structure
### Data Instances
Each task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.
There are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.
### Data Fields
'task': task identifier
'input': column elements of a specific row in the table.
'options': for multiple choice classification, it provides the options to choose from.
'output': target column element of the same row as input.
'pageTitle': the title of the page containing the table.
'outputColName': output column name
'url': url to the website containing the table
'wdcFile': WDC Web Table Corpus file
### Data Splits
The UnpredicTable datasets do not come with additional data splits.
## Dataset Creation
### Curation Rationale
Few-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.
### Source Data
#### Initial Data Collection and Normalization
We use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (http://webdatacommons.org/webtables/2015/EnglishStatistics.html). The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.
#### Who are the source language producers?
The dataset is extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/).
### Annotations
#### Annotation process
Manual annotation was only carried out for the [UnpredicTable-rated-low](https://huggingface.co/datasets/MicPie/unpredictable_rated-low),
[UnpredicTable-rated-medium](https://huggingface.co/datasets/MicPie/unpredictable_rated-medium), and [UnpredicTable-rated-high](https://huggingface.co/datasets/MicPie/unpredictable_rated-high) data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.
#### Who are the annotators?
Annotations were carried out by a lab assistant.
### Personal and Sensitive Information
The data was extracted from [WDC Web Table Corpora](http://webdatacommons.org/webtables/), which in turn extracted tables from the [Common Crawl](https://commoncrawl.org/). We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.
## Considerations for Using the Data
### Social Impact of Dataset
This dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.
### Discussion of Biases
Since our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.
### Other Known Limitations
No additional known limitations.
## Additional Information
### Dataset Curators
Jun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez
### Licensing Information
Apache 2.0
### Citation Information
```
@misc{chan2022few,
author = {Chan, Jun Shern and Pieler, Michael and Jao, Jonathan and Scheurer, Jérémy and Perez, Ethan},
title = {Few-shot Adaptation Works with UnpredicTable Data},
publisher={arXiv},
year = {2022},
url = {https://arxiv.org/abs/2208.01009}
}
```
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.