author stringlengths 2 29 ⌀ | cardData null | citation stringlengths 0 9.58k ⌀ | description stringlengths 0 5.93k ⌀ | disabled bool 1 class | downloads float64 1 1M ⌀ | gated bool 2 classes | id stringlengths 2 108 | lastModified stringlengths 24 24 | paperswithcode_id stringlengths 2 45 ⌀ | private bool 2 classes | sha stringlengths 40 40 | siblings list | tags list | readme_url stringlengths 57 163 | readme stringlengths 0 977k |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
bigscience-catalogue-data | null | null | null | false | null | false | bigscience-catalogue-data/qadi | 2022-01-28T15:16:48.000Z | null | true | 6e379178c5e1ab8d16df91e8fe5810b3db22b121 | [] | [
"arxiv:2005.06557"
] | https://huggingface.co/datasets/bigscience-catalogue-data/qadi/resolve/main/README.md | |
bigscience-catalogue-data | null | null | null | false | null | false | bigscience-catalogue-data/shamela | 2022-01-27T13:00:02.000Z | null | true | 7a0b02491e95bde4afa067639947bf5d3e6ebb77 | [] | [] | https://huggingface.co/datasets/bigscience-catalogue-data/shamela/resolve/main/README.md | |
bigscience-catalogue-data | null | null | null | false | null | false | bigscience-catalogue-data/urdu-monolingual-corpus | 2022-02-03T17:53:35.000Z | null | true | 32717642149f7e0a4d1ae1c080d00e1fb7e8fcbc | [] | [
"license:cc-by-nc-sa-3.0"
] | https://huggingface.co/datasets/bigscience-catalogue-data/urdu-monolingual-corpus/resolve/main/README.md | |
bigscience-catalogue-data | null | null | null | false | null | false | bigscience-catalogue-data/lm_en_s2orc_ai2_abstracts | 2022-02-18T15:27:59.000Z | null | true | cb2c0402fd2d2dc8c764220acf063514af68b47a | [] | [
"license:cc-by-nc-4.0"
] | https://huggingface.co/datasets/bigscience-catalogue-data/lm_en_s2orc_ai2_abstracts/resolve/main/README.md | |
bigscience-catalogue-data | null | null | null | false | null | false | bigscience-catalogue-data/lm_en_s2orc_ai2_pdf_parses | 2022-02-18T16:53:55.000Z | null | true | 45b0212282017441962e69443342f02522c8f4e4 | [] | [
"license:cc-by-4.0"
] | https://huggingface.co/datasets/bigscience-catalogue-data/lm_en_s2orc_ai2_pdf_parses/resolve/main/README.md | |
bigscience-catalogue-data | null | null | null | false | null | false | bigscience-catalogue-data/lm_fr_wikihow_human_instructions | 2022-02-21T18:40:07.000Z | null | true | abe98fc2f59683b2e0d7d732d360e423ea613eb4 | [] | [
"license:cc-by-nc-sa-4.0"
] | https://huggingface.co/datasets/bigscience-catalogue-data/lm_fr_wikihow_human_instructions/resolve/main/README.md | |
bigscience-catalogue-data | null | null | Leipzig Wortschatz Crawl | false | null | false | bigscience-catalogue-data/lm_indic-ur_leipzig_wortschatz_urdu-pk_web_2019_sentences | 2022-02-05T03:22:53.000Z | null | true | c1669fade21f7b9356d1a134b18b72c64a1b1d1b | [] | [
"license:other"
] | https://huggingface.co/datasets/bigscience-catalogue-data/lm_indic-ur_leipzig_wortschatz_urdu-pk_web_2019_sentences/resolve/main/README.md | |
bigscience-catalogue-data | null | null | Leipzig Wortschatz Crawl | false | null | false | bigscience-catalogue-data/lm_indic-ur_leipzig_wortschatz_urdu_newscrawl_2016_sentences | 2022-02-05T03:22:07.000Z | null | true | 6652b75c24393140edaa111816c952ea1f279f99 | [] | [
"license:other"
] | https://huggingface.co/datasets/bigscience-catalogue-data/lm_indic-ur_leipzig_wortschatz_urdu_newscrawl_2016_sentences/resolve/main/README.md | |
bigscience-catalogue-data | null | null | We release a sizeable monolingual Urdu corpus automatically tagged with part-of-speech tags. We extend the work of Jawaid and Bojar (2012) who use three different taggers and then apply a voting scheme to disambiguate among the different choices suggested by each tagger. We run this complex ensemble on a large monolingual corpus and release the both plain and tagged corpora. | false | null | false | bigscience-catalogue-data/lm_indic-ur_urdu-monolingual-corpus | 2022-02-05T03:23:47.000Z | null | true | f71f77f96ed747723b248890f70f40e0ac54d9d5 | [] | [
"license:cc-by-nc-sa-3.0"
] | https://huggingface.co/datasets/bigscience-catalogue-data/lm_indic-ur_urdu-monolingual-corpus/resolve/main/README.md | |
biu-nlp | null | @inproceedings{brook-weiss-etal-2021-qa,
title = "{QA}-Align: Representing Cross-Text Content Overlap by Aligning Question-Answer Propositions",
author = "Brook Weiss, Daniela and
Roit, Paul and
Klein, Ayal and
Ernst, Ori and
Dagan, Ido",
booktitle = "Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2021",
address = "Online and Punta Cana, Dominican Republic",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.emnlp-main.778",
pages = "9879--9894",
abstract = "Multi-text applications, such as multi-document summarization, are typically required to model redundancies across related texts. Current methods confronting consolidation struggle to fuse overlapping information. In order to explicitly represent content overlap, we propose to align predicate-argument relations across texts, providing a potential scaffold for information consolidation. We go beyond clustering coreferring mentions, and instead model overlap with respect to redundancy at a propositional level, rather than merely detecting shared referents. Our setting exploits QA-SRL, utilizing question-answer pairs to capture predicate-argument relations, facilitating laymen annotation of cross-text alignments. We employ crowd-workers for constructing a dataset of QA-based alignments, and present a baseline QA alignment model trained over our dataset. Analyses show that our new task is semantically challenging, capturing content overlap beyond lexical similarity and complements cross-document coreference with proposition-level links, offering potential use for downstream tasks.",
} | This dataset contains QA-Alignments - annotations of cross-text content overlap.
The task input is two sentences from two documents, roughly talking about the same event, along with their QA-SRL annotations
which capture verbal predicate-argument relations in question-answer format. The output is a cross-sentence alignment between sets of QAs which denote the same information.
See the paper for details: QA-Align: Representing Cross-Text Content Overlap by Aligning Question-Answer Propositions, Brook Weiss et. al., EMNLP 2021.
Here we provide both the QASRL annotations and the QA-Align annotations for the target sentences. | false | 321 | false | biu-nlp/qa_align | 2021-11-19T01:01:40.000Z | null | false | d8ed6544788253c114d64a2ba69c0b6c0f2ffa2d | [] | [] | https://huggingface.co/datasets/biu-nlp/qa_align/resolve/main/README.md | # QA-Align
This dataset contains QA-Alignments --- fine-grained annotations of cross-text content overlap.
The task input is two sentences from two documents, roughly talking about the same event, along with their QA-SRL annotations
which capture verbal predicate-argument relations in question-answer format. The output is a cross-sentence alignment between sets of QAs which denote the same information.
See the paper for details: [QA-Align: Representing Cross-Text Content Overlap by Aligning Question-Answer Propositions, Brook Weiss et. al., EMNLP 2021](https://aclanthology.org/2021.emnlp-main.778/).
The script downloads the data from the original [GitHub repository](https://github.com/DanielaBWeiss/QA-ALIGN).
### Format
The dataset contains the following important features:
* `abs_sent_id_1`, `abs_sent_id_2` - unique sentence ids, unique across all data sources.
* `text_1`, `text_2`, `prev_text_1`, `prev_text_2` - the two candidate sentences for alignments. The "prev" (previous) sentences are for context (shown to workers and for the model).
* `qas_1`, `qas_2` - the sets of QASRL QAs for each sentence. For test and dev they were created by workers, while in train, the QASRL parser generated them.
* `alignments` - the aligned QAs that workers have matched. This is the list of qa-alignments, where a single alignment looks like this:
```json
{'sent1': [{'qa_uuid': '33_1ecbplus~!~8~!~195~!~12~!~charged~!~4082',
'verb': 'charged',
'verb_idx': 12,
'question': 'Who was charged?',
'answer': 'the two youths',
'answer_range': '9:11'}],
'sent2': [{'qa_uuid': '33_8ecbplus~!~3~!~328~!~11~!~accused~!~4876',
'verb': 'accused',
'verb_idx': 11,
'question': 'Who was accused of something?',
'answer': 'two men',
'answer_range': '9:10'}]}
```
Where the for each sentence, we save a list of the aligned QAs from that sentence.
Note that this single alignment may contain multiple QAs for each sentence. While 96% of the data are one-to-one alignments, 4% contain many-to-many alignment (although most of the time it's a 2-to-1).
|
biu-nlp | null | @inproceedings{roit2020controlled,
title={Controlled Crowdsourcing for High-Quality QA-SRL Annotation},
author={Roit, Paul and Klein, Ayal and Stepanov, Daniela and Mamou, Jonathan and Michael, Julian and Stanovsky, Gabriel and Zettlemoyer, Luke and Dagan, Ido},
booktitle={Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics},
pages={7008--7013},
year={2020}
} | The dataset contains question-answer pairs to model verbal predicate-argument structure.
The questions start with wh-words (Who, What, Where, What, etc.) and contain a verb predicate in the sentence; the answers are phrases in the sentence.
This dataset, a.k.a "QASRL-GS" (Gold Standard) or "QASRL-2020", was constructed via controlled crowdsourcing.
See the paper for details: Controlled Crowdsourcing for High-Quality QA-SRL Annotation, Roit et. al., 2020 | false | 712 | false | biu-nlp/qa_srl2020 | 2022-10-17T20:49:01.000Z | null | false | 80e6b8ce552fc15f9ee698b414d677db1d6567fd | [] | [] | https://huggingface.co/datasets/biu-nlp/qa_srl2020/resolve/main/README.md | # QA-SRL 2020 (Gold Standard)
The dataset contains question-answer pairs to model verbal predicate-argument structure.
The questions start with wh-words (Who, What, Where, What, etc.) and contain a verb predicate in the sentence; the answers are phrases in the sentence.
This dataset, a.k.a "QASRL-GS" (Gold Standard) or "QASRL-2020", which was constructed via controlled crowdsourcing, includes high-quality QA-SRL annotations to serve as an evaluation set (dev and test) for models trained on the large-scale QA-SRL dataset (you can find it in this hub as [biu-nlp/qa_srl2018](https://huggingface.co/datasets/biu-nlp/qa_srl2018)).
See the paper for details: [Controlled Crowdsourcing for High-Quality QA-SRL Annotation, Roit et. al., 2020](https://aclanthology.org/2020.acl-main.626/).
Check out our [GitHub repository](https://github.com/plroit/qasrl-gs) to find code for evaluation.
The dataset was annotated by selected workers from Amazon Mechanical Turk. |
biu-nlp | null | @inproceedings{klein2020qanom,
title={QANom: Question-Answer driven SRL for Nominalizations},
author={Klein, Ayal and Mamou, Jonathan and Pyatkin, Valentina and Stepanov, Daniela and He, Hangfeng and Roth, Dan and Zettlemoyer, Luke and Dagan, Ido},
booktitle={Proceedings of the 28th International Conference on Computational Linguistics},
pages={3069--3083},
year={2020}
} | The dataset contains question-answer pairs to model predicate-argument structure of deverbal nominalizations.
The questions start with wh-words (Who, What, Where, What, etc.) and contain a the verbal form of a nominalization from the sentence;
the answers are phrases in the sentence.
See the paper for details: QANom: Question-Answer driven SRL for Nominalizations (Klein et. al., COLING 2020)
For previewing the QANom data along with the verbal annotations of QASRL, check out "https://browse.qasrl.org/".
This dataset was annotated by selected workers from Amazon Mechanical Turk. | false | 321 | false | biu-nlp/qanom | 2022-10-18T09:50:01.000Z | null | false | 5499db7c8223f09187bc8b4bc81d689758ceb5f8 | [] | [] | https://huggingface.co/datasets/biu-nlp/qanom/resolve/main/README.md | # QANom
This dataset contains question-answer pairs to model the predicate-argument structure of deverbal nominalizations.
The questions start with wh-words (Who, What, Where, What, etc.) and contain the verbal form of a nominalization from the sentence;
the answers are phrases in the sentence.
See the paper for details: [QANom: Question-Answer driven SRL for Nominalizations (Klein et. al., COLING 2020)](https://www.aclweb.org/anthology/2020.coling-main.274/)
For previewing the QANom data along with the verbal annotations of QASRL, check out https://browse.qasrl.org/.
Also check out our [GitHub repository](https://github.com/kleinay/QANom) to find code for nominalization identification, QANom annotation, evaluation, and models.
The dataset was annotated by selected workers from Amazon Mechanical Turk. |
blinoff | null | null | This dataset contains 190,335 Russian Q&A posts from a medical related forum. | false | 329 | false | blinoff/medical_qa_ru_data | 2022-07-02T06:24:13.000Z | null | false | b7c9fe729920578a60d6a294a6f6a81496d6c6fc | [] | [
"language:ru",
"license:unknown",
"multilinguality:monolingual",
"size_categories:100K<n<1M",
"source_datasets:original",
"task_categories:question-answering",
"task_ids:closed-domain-qa"
] | https://huggingface.co/datasets/blinoff/medical_qa_ru_data/resolve/main/README.md | ---
annotations_creators: []
language_creators: []
language:
- ru
license:
- unknown
multilinguality:
- monolingual
pretty_name: Medical Q&A Russian Data
size_categories:
- 100K<n<1M
source_datasets:
- original
task_categories:
- question-answering
task_ids:
- closed-domain-qa
---
### Dataset Summary
This dataset contains 190,335 Russian Q&A posts from a medical related forum.
### Dataset Fields
* date: date and time of the asked question, like '26 Октября 2018, 08:30'
* categ: question category
* theme: question topic
* desc: question text
* ans: question answers separated with ';\n'
* spec10: if present, one of 10 medical specializations
|
bs-modeling-metadata | null | null | null | false | 323 | false | bs-modeling-metadata/OSCAR_Entity_13_000 | 2021-09-15T14:20:53.000Z | null | false | 4949de27eaa80078f2253d1d254709a5e06a47a7 | [] | [] | https://huggingface.co/datasets/bs-modeling-metadata/OSCAR_Entity_13_000/resolve/main/README.md | The dataset is in the form of a json lines file with 10,657 examples, where an example consists of text (extracted from the first 13,000 rows of OSCAR unshuffled English dataset) and metadata fields (entities).
Structure of an example.
```
{
"text": "This is exactly the sort of article to raise the profile of the club around the Midlands. Very positive and really focusses on how the club has improved over a short period of time and the bright prospects for the future \n\"Oxford Town\" - professional as always at the Birmingham Mail. Not only is Oxford a city, but Oxford United are pretty recognisable name to anyone who has ever taken even a vague interest in English football.",
"metadata": [
{
"key": "entity",
"type": "local",
"char_start_idx": 80,
"char_end_idx": 88,
"value": "Midlands"
},
{
"key": "entity",
"type": "local",
"char_start_idx": 225,
"char_end_idx": 236,
"value": "Oxford Town"
},
{
"key": "entity",
"type": "local",
"char_start_idx": 270,
"char_end_idx": 285,
"value": "Birmingham_Mail"
},
{
"key": "entity",
"type": "local",
"char_start_idx": 299,
"char_end_idx": 305,
"value": "Oxford"
},
{
"key": "entity",
"type": "local",
"char_start_idx": 318,
"char_end_idx": 331,
"value": "Oxford_United_Stars_F.C."
},
{
"key": "entity",
"type": "local",
"char_start_idx": 415,
"char_end_idx": 422,
"value": "England"
}
]
}
```
|
bs-modeling-metadata | null | null | null | false | 322 | false | bs-modeling-metadata/website_metadata_c4 | 2021-11-24T14:04:30.000Z | null | false | f6cba351f9d1893a3b60f76455cdbd64fc0239c7 | [] | [] | https://huggingface.co/datasets/bs-modeling-metadata/website_metadata_c4/resolve/main/README.md | The dataset is in the form of a json lines file with 1,20,000 examples, where an example consists of text (extracted from C4 English dataset) and metadata fields (website description extracted from Wikipedia).
Example:
```
{
"text": "US10289222B2 - Handling of touch events in a browser environment - Google Patents\nHandling of touch events in a browser environment Download PDF\nUS10289222B2\nUS10289222B2 US13/857,848 US201313857848A US10289222B2 US 10289222 B2 US10289222 B2 US 10289222B2 US 201313857848 A US201313857848 A US 201313857848A US 10289222 B2 US10289222 B2 US 10289222B2\nUS13/857,848\nUS20130222244A1 (en\nEli Joshua FIDLER\nMichael Thomas Winkler\nMatthew Nicholaos STAIKOS\nJoseph Charles MASON\n2011-01-05 Priority to US12/985,337 priority Critical patent/US8438473B2/en\n2013-04-05 Application filed by BlackBerry Ltd filed Critical BlackBerry Ltd\n2013-04-05 Priority to US13/857,848 priority patent/US10289222B2/en\n2013-06-26 Assigned to RESEARCH IN MOTION CORPORATION reassignment RESEARCH IN MOTION CORPORATION ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: Winkler, Michael Thomas\n2013-06-26 Assigned to RESEARCH IN MOTION LIMITED reassignment RESEARCH IN MOTION LIMITED ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: Fidler, Eli Joshua, Mak, Genevieve Elizabeth, Mason, Joseph Charles, STAIKOS, MATTHEW\n2013-08-29 Publication of US20130222244A1 publication Critical patent/US20130222244A1/en\n2016-03-08 Assigned to RESEARCH IN MOTION LIMITED reassignment RESEARCH IN MOTION LIMITED ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: RESEARCH IN MOTION CORPORATION\n2019-05-14 Publication of US10289222B2 publication Critical patent/US10289222B2/en\nHandling of touch events in a browser environment is disclosed. An example method includes, while a document is displayed on a touchscreen display of a device, detecting a touch event at the touchscreen display, and selectively processing the detected touch event using one of a default hander, a touch event handler, and a conversion to one or more mouse events, according to a touch event handling property defined for the document.\nThe present application relates generally to the processing of detected user input events in a web browser.\nComputing devices such as desktop computers are typically equipped with external pointing devices, such as a mouse, to permit cursor-based user interaction with content executing on the computer.,
"metadata": [
{
"key": "website_description",
"type": "global",
"value": "Google Patents is a search engine from Google that indexes patents and patent applications."
}
]
}
``` |
bsc | null | AnCora Catalan NER.
This is a dataset for Named Eentity Reacognition (NER) from Ancora corpus adapted for
Machine Learning and Language Model evaluation purposes.
Since multiwords (including Named Entites) in the original Ancora corpus are aggregated as
a single lexical item using underscores (e.g. "Ajuntament_de_Barcelona")
we splitted them to align with word-per-line format, and added conventional Begin-Inside-Outside (IOB)
tags to mark and classify Named Entites.
We did not filter out the different categories of NEs from Ancora (weak and strong).
We did 6 minor edits by hand.
AnCora corpus is used under [CC-by] (https://creativecommons.org/licenses/by/4.0/) licence.
This dataset was developed by BSC TeMU as part of the AINA project, and to enrich the Catalan Language Understanding Benchmark (CLUB). | false | 321 | false | bsc/ancora-ca-ner | 2021-08-30T17:06:55.000Z | null | false | 8bba9af9375dcac303653a5c392420bf1a53756e | [] | [] | https://huggingface.co/datasets/bsc/ancora-ca-ner/resolve/main/README.md | # Named Entites from Ancora Corpus
<font size="+2">
<strong>
<span style="color:red">
WARNING:
</span>
</strong>
</font>
This repository is now superseded by [BSC-TeMU/ancora-ca-ner](https://huggingface.co/datasets/BSC-TeMU/ancora-ca-ner). Future updates will be released in the new repository, so it is highly recommended to load the dataset using the new path:
```python
from datasets import load_dataset
dataset = load_dataset("BSC-TeMU/ancora-ca-ner")
```
From now on, all models and datasets from the BSC's Text Mining Unit will be published on the [official organization account](https://huggingface.co/BSC-TeMU).
## License
<a rel="license" href="https://creativecommons.org/licenses/by/4.0/"><img alt="Attribution 4.0 International License" style="border-width:0" src="https://chriszabriskie.com/img/cc-by.png" width="100" /></a><br />This work is licensed under a <a rel="license" href="https://creativecommons.org/licenses/by/4.0/">Attribution 4.0 International License</a>. | |
bsc | null | Rodriguez-Penagos, Carlos Gerardo, Armentano-Oller, Carme, Gonzalez-Agirre, Aitor, & Gibert Bonet, Ona. (2021).
Semantic Textual Similarity in Catalan (Version 1.0.1) [Data set].
Zenodo. http://doi.org/10.5281/zenodo.4761434 | Semantic Textual Similarity in Catalan.
STS corpus is a benchmark for evaluating Semantic Text Similarity in Catalan.
It consists of more than 3000 sentence pairs, annotated with the semantic similarity between them,
using a scale from 0 (no similarity at all) to 5 (semantic equivalence).
It is done manually by 4 different annotators following our guidelines based on previous work from the SemEval challenges (https://www.aclweb.org/anthology/S13-1004.pdf).
The source data are scraped sentences from the Catalan Textual Corpus (https://doi.org/10.5281/zenodo.4519349), used under CC-by-SA-4.0 licence (https://creativecommons.org/licenses/by-sa/4.0/). The dataset is released under the same licence.
This dataset was developed by BSC TeMU as part of the AINA project, and to enrich the Catalan Language Understanding Benchmark (CLUB).
This is the version 1.0.2 of the dataset with the complete human and automatic annotations and the analysis scripts. It also has a more accurate license.
This dataset can be used to build and score semantic similiarity models. | false | 324 | false | bsc/sts-ca | 2021-08-30T17:20:03.000Z | null | false | 724c570fd063a86bc0f6cf2d7be9dc4da0ded036 | [] | [] | https://huggingface.co/datasets/bsc/sts-ca/resolve/main/README.md | # Semantic Textual Similarity in Catalan
<font size="+2">
<strong>
<span style="color:red">
WARNING:
</span>
</strong>
</font>
This repository is now superseded by [BSC-TeMU/sts-ca](https://huggingface.co/datasets/BSC-TeMU/sts-ca). Future updates will be released in the new repository, so it is highly recommended to load the dataset using the new path:
```python
from datasets import load_dataset
dataset = load_dataset("BSC-TeMU/sts-ca")
```
From now on, all models and datasets from the BSC's Text Mining Unit will be published on the [official organization account](https://huggingface.co/BSC-TeMU).
## License
<a rel="license" href="https://creativecommons.org/licenses/by-sa/4.0/"><img alt="Attribution-ShareAlike 4.0 International License" style="border-width:0" src="https://i.creativecommons.org/l/by/4.0/88x31.png" width="100" /></a><br />This work is licensed under a <a rel="license" href="https://creativecommons.org/licenses/by-sa/4.0/">Attribution-ShareAlike 4.0 International License</a>. |
bsc | null | Carrino, Casimiro Pio, Rodriguez-Penagos, Carlos Gerardo, & Armentano-Oller, Carme. (2021).
TeCla: Text Classification Catalan dataset (Version 1.0) [Data set].
Zenodo. http://doi.org/10.5281/zenodo.4627198 | TeCla: Text Classification Catalan dataset
Catalan News corpus for Text classification, crawled from ACN (Catalan News Agency) site: www.acn.cat
Corpus de notícies en català per a classificació textual, extret del web de l'Agència Catalana de Notícies - www.acn.cat | false | 322 | false | bsc/tecla | 2021-08-30T17:12:59.000Z | null | false | 5c73e8058b6676acaa79f5a370c70c0ca2626b17 | [] | [] | https://huggingface.co/datasets/bsc/tecla/resolve/main/README.md | # TeCla (Text Classification) Catalan dataset
<font size="+2">
<strong>
<span style="color:red">
WARNING:
</span>
</strong>
</font>
This repository is now superseded by [BSC-TeMU/tecla](https://huggingface.co/datasets/BSC-TeMU/tecla). Future updates will be released in the new repository, so it is highly recommended to load the dataset using the new path:
```python
from datasets import load_dataset
dataset = load_dataset("BSC-TeMU/tecla")
```
From now on, all models and datasets from the BSC's Text Mining Unit will be published on the [official organization account](https://huggingface.co/BSC-TeMU).
## License
<a rel="license" href="https://creativecommons.org/licenses/by-nc-nd/4.0/"><img alt="Attribution-NonCommercial-NoDerivatives 4.0 International License" style="border-width:0" src="http://d2klr1ixr44jla.cloudfront.net/306/125/0.5-0.5/assets/images/55132bfeb13b7b027c000041.png" width="100" /></a><br />This work is licensed under a <a rel="license" href="https://creativecommons.org/licenses/by-nc-nd/4.0/">Attribution-NonCommercial-NoDerivatives 4.0 International License</a>. |
bsc | null | Rodriguez-Penagos, Carlos Gerardo, & Armentano-Oller, Carme. (2021).
ViquiQuAD: an extractive QA dataset from Catalan Wikipedia (Version ViquiQuad_v.1.0.1)
[Data set]. Zenodo. http://doi.org/10.5281/zenodo.4761412 | ViquiQuAD: an extractive QA dataset from Catalan Wikipedia.
This dataset contains 3111 contexts extracted from a set of 597 high quality original (no translations)
articles in the Catalan Wikipedia "Viquipèdia" (ca.wikipedia.org), and 1 to 5 questions with their
answer for each fragment. Viquipedia articles are used under CC-by-sa licence.
This dataset can be used to build extractive-QA and Language Models.
Funded by the Generalitat de Catalunya, Departament de Polítiques Digitals i Administració Pública (AINA),
MT4ALL and Plan de Impulso de las Tecnologías del Lenguaje (Plan TL). | false | 326 | false | bsc/viquiquad | 2021-08-30T17:22:59.000Z | null | false | d7f76452359631f2be589609060b6de8eea90428 | [] | [] | https://huggingface.co/datasets/bsc/viquiquad/resolve/main/README.md | # ViquiQuAD, An extractive QA dataset for catalan, from the Wikipedia
<font size="+2">
<strong>
<span style="color:red">
WARNING:
</span>
</strong>
</font>
This repository is now superseded by [BSC-TeMU/viquiquad](https://huggingface.co/datasets/BSC-TeMU/viquiquad). Future updates will be released in the new repository, so it is highly recommended to load the dataset using the new path:
```python
from datasets import load_dataset
dataset = load_dataset("BSC-TeMU/viquiquad")
```
From now on, all models and datasets from the BSC's Text Mining Unit will be published on the [official organization account](https://huggingface.co/BSC-TeMU).
## License
<a rel="license" href="https://creativecommons.org/licenses/by-sa/4.0/"><img alt="Attribution-ShareAlike 4.0 International License" style="border-width:0" src="https://i.creativecommons.org/l/by/4.0/88x31.png" width="100"/></a><br />This work is licensed under a <a rel="license" href="https://creativecommons.org/licenses/by-sa/4.0/">Attribution-ShareAlike 4.0 International License</a>.
|
bsc | null | Carlos Gerardo Rodriguez-Penagos, & Carme Armentano-Oller. (2021). XQuAD-ca [Data set].
Zenodo. http://doi.org/10.5281/zenodo.4757559 | Professional translation into Catalan of XQuAD dataset (https://github.com/deepmind/xquad).
XQuAD (Cross-lingual Question Answering Dataset) is a benchmark dataset for evaluating
cross-lingual question answering performance.
The dataset consists of a subset of 240 paragraphs and 1190 question-answer pairs from
the development set of SQuAD v1.1 (Rajpurkar et al., 2016) together with
their professional translations into ten languages:
Spanish, German, Greek, Russian, Turkish, Arabic, Vietnamese, Thai, Chinese, and Hindi.
Rumanian was added later.
We added the 13th language to the corpus using also professional native catalan translators.
XQuAD and XQuAD-Ca datasets are released under CC-by-sa licence. | false | 323 | false | bsc/xquad-ca | 2021-08-30T17:16:31.000Z | null | false | 34dc222345ef8a4ed8ae0a4d181f08c9e8cc548b | [] | [] | https://huggingface.co/datasets/bsc/xquad-ca/resolve/main/README.md | # XQuAD-Ca
<font size="+2">
<strong>
<span style="color:red">
WARNING:
</span>
</strong>
</font>
This repository is now superseded by [BSC-TeMU/xquad-ca](https://huggingface.co/datasets/BSC-TeMU/xquad-ca). Future updates will be released in the new repository, so it is highly recommended to load the dataset using the new path:
```python
from datasets import load_dataset
dataset = load_dataset("BSC-TeMU/xquad-ca")
```
From now on, all models and datasets from the BSC's Text Mining Unit will be published on the [official organization account](https://huggingface.co/BSC-TeMU).
## License
<a rel="license" href="https://creativecommons.org/licenses/by-sa/4.0/"><img alt="Attribution-ShareAlike 4.0 International License" style="border-width:0" src="https://i.creativecommons.org/l/by/4.0/88x31.png" width="100" /></a><br />This work is licensed under a <a rel="license" href="https://creativecommons.org/licenses/by-sa/4.0/">Attribution-ShareAlike 4.0 International License</a>. |
caca | null | null | null | false | 166 | false | caca/zscczs | 2021-04-07T09:02:09.000Z | null | false | a77e6b9e25050d202bc69d78b3cdd9529ef10029 | [] | [] | https://huggingface.co/datasets/caca/zscczs/resolve/main/README.md | |
cahya | null | null | null | false | 166 | false | cahya/persona_empathetic | 2022-02-19T22:49:35.000Z | null | false | 47de2aba7b0adf7b8c37a568df2d4b69717d8dcb | [] | [
"license:mit"
] | https://huggingface.co/datasets/cahya/persona_empathetic/resolve/main/README.md | ---
license: mit
---
|
cakiki | null | @dataset{yamen_ajjour_2020_4139439,
author = {Yamen Ajjour and
Henning Wachsmuth and
Johannes Kiesel and
Martin Potthast and
Matthias Hagen and
Benno Stein},
title = {args.me corpus},
month = oct,
year = 2020,
publisher = {Zenodo},
version = {1.0-cleaned},
doi = {10.5281/zenodo.4139439},
url = {https://doi.org/10.5281/zenodo.4139439}
} | The args.me corpus (version 1.0, cleaned) comprises 382 545 arguments crawled from four debate portals in the middle of 2019. The debate portals are Debatewise, IDebate.org, Debatepedia, and Debate.org. The arguments are extracted using heuristics that are designed for each debate portal. | false | 640 | false | cakiki/args_me | 2022-10-25T09:07:25.000Z | null | false | da29f2b2fc7c86176813b8a6440f73e0823f05d3 | [] | [
"annotations_creators:machine-generated",
"language_creators:crowdsourced",
"language:'en-US'",
"license:cc-by-4.0",
"multilinguality:monolingual",
"size_categories:100K<n<1M",
"source_datasets:original",
"task_categories:text-retrieval",
"task_ids:document-retrieval"
] | https://huggingface.co/datasets/cakiki/args_me/resolve/main/README.md | ---
annotations_creators:
- machine-generated
language_creators:
- crowdsourced
language:
- '''en-US'''
license:
- cc-by-4.0
multilinguality:
- monolingual
pretty_name: Webis args.me argument corpus
size_categories:
- 100K<n<1M
source_datasets:
- original
task_categories:
- text-retrieval
task_ids:
- document-retrieval
---
# Dataset Card for the args.me corpus
## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Dataset Usage](#dataset-usage)
- [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://zenodo.org/record/4139439
- **Repository:** https://git.webis.de/code-research/arguana/args/args-framework
- **Paper:** [Building an Argument Search Engine for the Web](https://webis.de/downloads/publications/papers/wachsmuth_2017f.pdf)
- **Leaderboard:** https://touche.webis.de/
- **Point of Contact:** [Webis Group](https://webis.de/people.html)
### Dataset Summary
The args.me corpus (version 1.0, cleaned) comprises 382 545 arguments crawled from four debate portals in the middle of 2019. The debate portals are Debatewise, IDebate.org, Debatepedia, and Debate.org. The arguments are extracted using heuristics that are designed for each debate portal.
### Dataset Usage
```python
import datasets
args = datasets.load_dataset('cakiki/args_me', 'corpus', streaming=True)
args_iterator = iter(args)
for arg in args_iterator:
print(args['conclusion'])
print(args['id'])
print(args['argument'])
print(args['stance'])
break
```
### Supported Tasks and Leaderboards
Document Retrieval, Argument Retrieval for Controversial Questions
### Languages
The args.me corpus is monolingual; it only includes English (mostly en-US) documents.
## Dataset Structure
### Data Instances
#### Corpus
```
{'conclusion': 'Science is the best!',
'id': 'd6517702-2019-04-18T12:36:24Z-00000-000',
'argument': 'Science is aright I guess, but Physical Education (P.E) is better. Think about it, you could sit in a classroom for and hour learning about molecular reconfiguration, or you could play football with your mates. Why would you want to learn about molecular reconfiguration anyway? I think the argument here would be based on, healthy mind or healthy body. With science being the healthy mind and P.E being the healthy body. To work this one out all you got to do is ask Steven Hawkins. Only 500 words',
'stance': 'CON'}
```
### 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
[Creative Commons Attribution 4.0 International (CC BY 4.0)](https://creativecommons.org/licenses/by/4.0/)
### Citation Information
```
@dataset{yamen_ajjour_2020_4139439,
author = {Yamen Ajjour and
Henning Wachsmuth and
Johannes Kiesel and
Martin Potthast and
Matthias Hagen and
Benno Stein},
title = {args.me corpus},
month = oct,
year = 2020,
publisher = {Zenodo},
version = {1.0-cleaned},
doi = {10.5281/zenodo.4139439},
url = {https://doi.org/10.5281/zenodo.4139439}
}
```
|
cakiki | null | null | null | false | 167 | false | cakiki/arxiv-metadata | 2022-02-03T20:57:23.000Z | null | false | 3a0dac229b4e21cbde67cb06af07d11fd8bb7c75 | [] | [
"license:cc0-1.0"
] | https://huggingface.co/datasets/cakiki/arxiv-metadata/resolve/main/README.md | ---
license: cc0-1.0
---
|
cakiki | null | null | null | false | 325 | false | cakiki/en_wiki_quote | 2022-02-03T17:36:03.000Z | null | false | 0882808a91679e12e98407691dabd28115bff670 | [] | [
"license:cc-by-sa-3.0"
] | https://huggingface.co/datasets/cakiki/en_wiki_quote/resolve/main/README.md | ---
license: cc-by-sa-3.0
---
|
caltonji | null | null | null | false | 327 | false | caltonji/harrypotter_squad_v2_2 | 2021-12-31T20:01:23.000Z | null | false | aba728f708a01b22a508061490cf389dd15f6ca2 | [] | [] | https://huggingface.co/datasets/caltonji/harrypotter_squad_v2_2/resolve/main/README.md | ## Dataset Summary
Contains 15 Harry Potter trivia questions in Squadv2 format, 3 of which are unanswerable.
## Model Performance
[Test Notebook](https://colab.research.google.com/drive/1VFUJKV7eun68XgQDAHSHsbvoM_CGHzWA?usp=sharing)
| Model | exact | f1 |
| ----------- | ----------- | ----------- |
| Albert Base ([twmkn9/albert-base-v2-squad2](https://huggingface.co/twmkn9/albert-base-v2-squad2)) | 46.6667 | 46.6667 |
| Albert XXLarge ([ahotrod/albert_xxlargev1_squad2_512](https://huggingface.co/ahotrod/albert_xxlargev1_squad2_512)) | 66.6667 | 66.6667 |
|
cassandra-themis | null | null | QR-AN Dataset: a classification dataset on french Parliament debates
This is a dataset for theme/topic classification, made of questions and answers from https://www2.assemblee-nationale.fr/recherche/resultats_questions.
It contains 188 unbalanced classes, 80k questions-answers divided into 3 splits: train (60k), val (10k) and test (10k). | false | 799 | false | cassandra-themis/QR-AN | 2022-10-24T20:31:22.000Z | null | false | 1059be355e830b808093595856135651e770d22c | [] | [
"language:fr",
"size_categories:10K<n<100K",
"task_categories:summarization",
"task_categories:text-classification",
"task_categories:text-generation",
"task_ids:multi-class-classification",
"task_ids:topic-classification",
"tags:conditional-text-generation"
] | https://huggingface.co/datasets/cassandra-themis/QR-AN/resolve/main/README.md | ---
language:
- fr
size_categories: 10K<n<100K
task_categories:
- summarization
- text-classification
- text-generation
task_ids:
- multi-class-classification
- topic-classification
tags:
- conditional-text-generation
---
**QR-AN Dataset: a classification and generation dataset of french Parliament questions-answers.**
This is a dataset for theme/topic classification, made of questions and answers from https://www2.assemblee-nationale.fr/recherche/resultats_questions . \
It contains 188 unbalanced classes, 80k questions-answers divided into 3 splits: train (60k), val (10k) and test (10k). \
Can be used for generation with 'qran_generation'
This dataset is compatible with the [`run_summarization.py`](https://github.com/huggingface/transformers/tree/master/examples/pytorch/summarization) script from Transformers if you add this line to the `summarization_name_mapping` variable:
```python
"ccdv/cass-summarization": ("question", "answer")
```
Compatible with [run_glue.py](https://github.com/huggingface/transformers/tree/master/examples/pytorch/text-classification) script:
```
export MODEL_NAME=camembert-base
export MAX_SEQ_LENGTH=512
python run_glue.py \
--model_name_or_path $MODEL_NAME \
--dataset_name cassandra-themis/QR-AN \
--do_train \
--do_eval \
--max_seq_length $MAX_SEQ_LENGTH \
--per_device_train_batch_size 8 \
--gradient_accumulation_steps 4 \
--learning_rate 2e-5 \
--num_train_epochs 1 \
--max_eval_samples 500 \
--output_dir tmp/QR-AN
``` |
castorini | null | @inproceedings{ogueji-etal-2021-small,
title = "Small Data? No Problem! Exploring the Viability of Pretrained Multilingual Language Models for Low-resourced Languages",
author = "Ogueji, Kelechi and
Zhu, Yuxin and
Lin, Jimmy",
booktitle = "Proceedings of the 1st Workshop on Multilingual Representation Learning",
month = nov,
year = "2021",
address = "Punta Cana, Dominican Republic",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.mrl-1.11",
pages = "116--126",
} | Corpus used for training AfriBERTa models | false | 1,812 | false | castorini/afriberta-corpus | 2022-10-19T21:33:04.000Z | null | false | d83da9653ef2a5f823c3693a28018e3009464522 | [] | [
"language:om",
"language:am",
"language:rw",
"language:rn",
"language:ha",
"language:ig",
"language:pcm",
"language:so",
"language:sw",
"language:ti",
"language:yo",
"language:multilingual",
"license:apache-2.0",
"task_categories:text-generation",
"task_ids:language-modeling"
] | https://huggingface.co/datasets/castorini/afriberta-corpus/resolve/main/README.md | ---
language:
- om
- am
- rw
- rn
- ha
- ig
- pcm
- so
- sw
- ti
- yo
- multilingual
license: apache-2.0
task_categories:
- text-generation
task_ids:
- language-modeling
---
# Dataset Card for AfriBERTa's Corpus
## Table of Contents
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Loading Dataset](#loading-dataset)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)
- [Data Splits](#data-splits)
- [Considerations for Using the Data](#considerations-for-using-the-data)
- [Discussion of Biases](#discussion-of-biases)
- [Additional Information](#additional-information)
- [Citation Information](#citation-information)
- [Contributions](#contributions)
## Dataset Description
### Dataset Summary
This is the corpus on which AfriBERTa was trained on.
The dataset is mostly from the BBC news website, but some languages also have data from Common Crawl.
- **Homepage:** https://github.com/keleog/afriberta
- **Models:**
- https://huggingface.co/castorini/afriberta_small
- https://huggingface.co/castorini/afriberta_base
- https://huggingface.co/castorini/afriberta_large
- **Paper:** https://aclanthology.org/2021.mrl-1.11/
- **Point of Contact:** kelechi.ogueji@uwaterloo.ca
### Supported Tasks and Leaderboards
The AfriBERTa corpus was mostly intended to pre-train language models.
### Languages
```
afaanoromoo
amharic
gahuza
hausa
igbo
pidgin
somali
swahili
tigrinya
yoruba
```
### Loading Dataset
An example to load the train split of the Somali corpus:
```
dataset = load_dataset("castorini/afriberta-corpus", "somali", split="train")
```
An example to load the test split of the Pidgin corpus:
```
dataset = load_dataset("castorini/afriberta-corpus", "pidgin", split="test")
```
## Dataset Structure
### Data Instances
Each data point is a line of text.
An example from the `igbo` dataset:
```
{"id": "6", "text": "Ngwá ọrụ na-echebe ma na-ebuli gị na kọmputa."}
```
### Data Fields
The data fields are:
- id: id of the example
- text: content as a string
### Data Splits
Each language has a train and test split, with varying sizes.
## Considerations for Using the Data
### Discussion of Biases
Since majority of the data is obtained from the BBC's news website, models trained on this dataset are likely going to
be biased towards the news domain.
Also, since some of the data is obtained from Common Crawl, care should be taken (especially for text generation models) since personal and sensitive information might be present.
## Additional Information
### Citation Information
```
@inproceedings{ogueji-etal-2021-small,
title = "Small Data? No Problem! Exploring the Viability of Pretrained Multilingual Language Models for Low-resourced Languages",
author = "Ogueji, Kelechi and
Zhu, Yuxin and
Lin, Jimmy",
booktitle = "Proceedings of the 1st Workshop on Multilingual Representation Learning",
month = nov,
year = "2021",
address = "Punta Cana, Dominican Republic",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.mrl-1.11",
pages = "116--126",
}
```
### Contributions
Thanks to [Kelechi Ogueji](https://github.com/keleog) for adding this dataset. |
castorini | null | null | null | false | 2,411 | false | castorini/mr-tydi-corpus | 2022-10-12T20:25:51.000Z | null | false | 3a3aa212bbe94a8cc0dc858710a3dad49d532054 | [] | [
"language:ar",
"language:bn",
"language:en",
"language:fi",
"language:id",
"language:ja",
"language:ko",
"language:ru",
"language:sw",
"language:te",
"language:th",
"multilinguality:multilingual",
"task_categories:text-retrieval",
"license:apache-2.0"
] | https://huggingface.co/datasets/castorini/mr-tydi-corpus/resolve/main/README.md | ---
language:
- ar
- bn
- en
- fi
- id
- fi
- ja
- ko
- ru
- sw
- te
- th
multilinguality:
- multilingual
task_categories:
- text-retrieval
license: apache-2.0
---
# Dataset Summary
Mr. TyDi is a multi-lingual benchmark dataset built on TyDi, covering eleven typologically diverse languages. It is designed for monolingual retrieval, specifically to evaluate ranking with learned dense representations.
This dataset stores documents of Mr. TyDi. To access the queries and judgments, please refer to [castorini/mr-tydi](https://huggingface.co/datasets/castorini/mr-tydi).
# Dataset Structure
The only configuration here is the `language`. As all three folds (train, dev and test) share the same corpus, there is only one fold 'train' under each language, unlike [castorini/mr-tydi](https://huggingface.co/datasets/castorini/mr-tydi).
An example of document data entry looks as follows:
```
{
'docid': '25#0',
'title': 'Autism',
'text': 'Autism is a developmental disorder characterized by difficulties with social interaction and communication, ...'
}
```
# Load Dataset
An example to load the dataset:
```
language = 'english'
dataset = load_dataset('castorini/mr-tydi-corpus', language, 'train')
```
# Citation Information
```
@article{mrtydi,
title={{Mr. TyDi}: A Multi-lingual Benchmark for Dense Retrieval},
author={Xinyu Zhang and Xueguang Ma and Peng Shi and Jimmy Lin},
year={2021},
journal={arXiv:2108.08787},
}
``` |
castorini | null | null | null | false | 2,316 | false | castorini/mr-tydi | 2022-10-12T20:25:19.000Z | null | false | 1d43c80218d06d0ef80f5b172ccabd848b948bc1 | [] | [
"language:ar",
"language:bn",
"language:en",
"language:fi",
"language:id",
"language:ja",
"language:ko",
"language:ru",
"language:sw",
"language:te",
"language:th",
"multilinguality:multilingual",
"task_categories:text-retrieval",
"license:apache-2.0"
] | https://huggingface.co/datasets/castorini/mr-tydi/resolve/main/README.md | ---
language:
- ar
- bn
- en
- fi
- id
- fi
- ja
- ko
- ru
- sw
- te
- th
multilinguality:
- multilingual
task_categories:
- text-retrieval
license: apache-2.0
---
# Dataset Summary
Mr. TyDi is a multi-lingual benchmark dataset built on TyDi, covering eleven typologically diverse languages. It is designed for monolingual retrieval, specifically to evaluate ranking with learned dense representations.
This dataset stores the queries, judgements, and example training data of Mr. TyDi. To access the corpus, please refer to [castorini/mr-tydi-corpus](https://huggingface.co/datasets/castorini/mr-tydi-corpus).
# Dataset Structure
The only configuration here is the `language`,
For each language, there are three splits: `train`, `dev`, and `test`.
The negative examples from training set are sampled from the top-30 BM25 runfiles on each language.
Specifically, we combine the **training** data for all languages under the `combined` configuration.
An example of `train` set looks as follows:
```
{
'query_id': '1',
'query': 'When was quantum field theory developed?',
'positive_passages': [
{
'docid': '25267#12',
'title': 'Quantum field theory',
'text': 'Quantum field theory naturally began with the study of electromagnetic interactions, as the electromagnetic field was the only known classical field as of the 1920s.'
},
...
]
'negative_passages': [
{
'docid': '346489#8',
'title': 'Local quantum field theory',
'text': 'More recently, the approach has been further implemented to include an algebraic version of quantum field ...'
},
...
],
}
```
An example of `dev` and `test` set looks as follows. We only provide the docid of positive passages here to save the space.
Also no candidate passages are provided at this point.
Note that to perform the retrieval, it need to be used together with [castorini/mr-tydi-corpus](https://huggingface.co/datasets/castorini/mr-tydi-corpus)
```
{
'query_id': '0',
'query': 'Is Creole a pidgin of French?',
'positive_passages': [
{
'docid': '3716905#1',
'title': '',
'text': ''
},
...
]
}
```
# Load Dataset
An example to load the dataset:
```
language = 'english'
# to load all train, dev and test sets
dataset = load_dataset('castorini/mr-tydi', language)
# or to load a specific set:
set_name = 'train'
dataset = load_dataset('castorini/mr-tydi', language, set_name)
```
Note that the 'combined' option has only the 'train' set.
# Citation Information
```
@article{mrtydi,
title={{Mr. TyDi}: A Multi-lingual Benchmark for Dense Retrieval},
author={Xinyu Zhang and Xueguang Ma and Peng Shi and Jimmy Lin},
year={2021},
journal={arXiv:2108.08787},
}
```
|
castorini | null | null | null | false | 323 | false | castorini/msmarco_v1_doc_doc2query-t5_expansions | 2022-07-02T19:16:12.000Z | null | false | 73205571221e7eac6953ed884e05c8625e06272c | [] | [
"language:en",
"license:apache-2.0"
] | https://huggingface.co/datasets/castorini/msmarco_v1_doc_doc2query-t5_expansions/resolve/main/README.md | ---
language:
- en
license: apache-2.0
---
# Dataset Summary
The repo provides queries generated for the MS MARCO V1 document corpus with docTTTTTquery (sometimes written as docT5query or doc2query-T5), the latest version of the doc2query family of document expansion models. The basic idea is to train a model, that when given an input document, generates questions that the document might answer (or more broadly, queries for which the document might be relevant). These predicted questions (or queries) are then appended to the original documents, which are then indexed as before. The docTTTTTquery model gets its name from the use of T5 as the expansion model.
# Dataset Structure
All three folds (train, dev and test) share the same corpus.
An example data entry looks as follows:
```
{
"id": "D1555982",
"predicted_queries": ["when find radius of star r", "what is r radius", "how to find out radius of star", "what is radius r", "what is radius of r", "how do you find radius of star igel", "which law states that radiation is proportional to radiation?", "what is the radius of a spherical star", "what is the radius of the star", "what is radius of star", "which radiation is produced during a solar radiation experiment?", "how to find radius r", "what is radius r of a star", "the hot glowing surfaces of stars emit energy in the form of", "what is the radius of a star", "what is the radius of a star", "how to find radius r on a star", "how to find radius r in a solar cell", "what kind of energy does a hot glowing surface of a star emit?", "what kind of energy does the hot glowing surface of stars emit"]
}
```
# Load Dataset
An example to load the dataset:
```
dataset = load_dataset('castorini/msmarco_v1_doc_doc2query-t5_expansions')
```
# Citation Information
```
@article{docTTTTTquery,
title={From doc2query to {docTTTTTquery}},
author={Nogueira, Rodrigo and Lin, Jimmy},
year={2019}
}
@article{emdt5,
author = "Ronak Pradeep and Rodrigo Nogueira and Jimmy Lin",
title = "The Expando-Mono-Duo Design Pattern for Text Ranking with Pretrained Sequence-to-Sequence Models",
journal = "arXiv:2101.05667",
year = 2021,
}
|
castorini | null | null | null | false | 323 | false | castorini/msmarco_v1_doc_segmented_doc2query-t5_expansions | 2021-11-10T04:51:35.000Z | null | false | 4254f0bda2a6e562cb2e53001220e0f1f981d2b8 | [] | [
"language:English",
"license:Apache License 2.0"
] | https://huggingface.co/datasets/castorini/msmarco_v1_doc_segmented_doc2query-t5_expansions/resolve/main/README.md | ---
language:
- English
license: "Apache License 2.0"
---
# Dataset Summary
The repo provides queries generated for the MS MARCO V1 document segmented corpus with docTTTTTquery (sometimes written as docT5query or doc2query-T5), the latest version of the doc2query family of document expansion models. The basic idea is to train a model, that when given an input document, generates questions that the document might answer (or more broadly, queries for which the document might be relevant). These predicted questions (or queries) are then appended to the original documents, which are then indexed as before. The docTTTTTquery model gets its name from the use of T5 as the expansion model.
# Dataset Structure
All three folds (train, dev and test) share the same corpus.
An example data entry looks as follows:
```
{
"id": "D1555982#0", "predicted_queries": ["when find radius of star r", "what is r radius", "how to find out radius of star", "what is radius r", "what is radius of r", "how do you find radius of star igel", "which law states that radiation is proportional to radiation?", "what is the radius of a spherical star", "what is the radius of the star", "what is radius of star"]
}
```
# Load Dataset
An example to load the dataset:
```
dataset = load_dataset('castorini/msmarco_v1_doc_segmented_doc2query-t5_expansions')
```
# Citation Information
```
@article{docTTTTTquery,
title={From doc2query to {docTTTTTquery}},
author={Nogueira, Rodrigo and Lin, Jimmy},
year={2019}
}
@article{emdt5,
author = "Ronak Pradeep and Rodrigo Nogueira and Jimmy Lin",
title = "The Expando-Mono-Duo Design Pattern for Text Ranking with Pretrained Sequence-to-Sequence Models",
journal = "arXiv:2101.05667",
year = 2021,
}
|
castorini | null | null | null | false | 326 | false | castorini/msmarco_v1_passage_doc2query-t5_expansions | 2022-06-21T17:45:43.000Z | null | false | aca81f4eabebd63c46026565b9123b17269bb1c4 | [] | [
"language:English",
"license:Apache License 2.0"
] | https://huggingface.co/datasets/castorini/msmarco_v1_passage_doc2query-t5_expansions/resolve/main/README.md | ---
language:
- English
license: "Apache License 2.0"
---
# Dataset Summary
The repo provides queries generated for the MS MARCO V1 passage corpus with docTTTTTquery (sometimes written as docT5query or doc2query-T5), the latest version of the doc2query family of document expansion models. The basic idea is to train a model, that when given an input document, generates questions that the document might answer (or more broadly, queries for which the document might be relevant). These predicted questions (or queries) are then appended to the original documents, which are then indexed as before. The docTTTTTquery model gets its name from the use of T5 as the expansion model.
# Dataset Structure
All three folds (train, dev and test) share the same corpus. The queries are generated from this corpus.
An example data entry looks as follows:
```
{
"id": "0",
"predicted_queries": ["what was important to the success of the manhattan project", "why was the manhattan project important?", "what was important about the manhattan project", "why was the success of the manhattan project so important?", "who was the manhattan project a scientific project for", "what was the manhattan project important for", "why was the manhattan project a success", "how was the success of the manhattan project", "why was the manhattan project important to the success of the project?", "what is the importance of communication amongst scientific minds", "what was the importance of scientific communication for the success of the manhattan project", "what was the purpose of the manhattan project", "why was the manhattan project significant?", "why was the manhattan project important", "why did scientists believe in atomic power", "why did scientists and engineers have to communicate?", "why was the manhattan project a success", "what was the purpose of the manhattan project", "why did scientists and engineers want to be involved in the manhattan project", "why are the scientists so valuable", "which of the following was an important outcome of the manhattan project?", "why was the manhattan project successful", "why was the manhattan project an important scientific achievement", "what was the success of manhattan", "what was the result of the manhattan project", "why was communications important to the success of the manhattan project?", "why the manhattan project was important", "why is it important to know who is the manhattan project", "what was the most important accomplishment to the success of the manhattan project?", "why was the manhattan project an important achievement?", "why was the manhattan project important to the success of the atomic bomb", "how did the manhattan project impact scientists?", "what were the effects of the manhattan project", "what were the results of the manhattan project and how did they affect the public", "what was the manhattan project", "why did scientists contribute to the success of the manhattan project", "why was communication important in the manhattan project", "what was the effect of the manhattan project on the world", "what was the importance of communication in the success of the manhattan project?", "why was communications important to the success of the manhattan project?", "why was the manhattan project important", "what was the manhattan project", "why was the success of the manhattan project important", "why was manhattan project a success", "what was important about the manhattan project", "what benefited from the success of the new york nuclear bomb", "what was the significance to the success of the manhattan project?", "why is communication important", "why was the manhattan project an important achievement", "why did the manhattan project work", "what was the manhattan project's success", "what was the significance of the manhattan experiment", "how important was communication to the success of the manhattan project", "why is communication important to the success of the manhattan project?", "what was the importance of the manhattan project", "why did scientists believe the manhattan project had the greatest impact on science?", "what was a critical effect of the manhattan project?", "why did the manhattan project succeed", "what was the importance of the manhattan project", "why was the manhattan project important", "why was the manhattan project a success?", "what was the importance of communication and communication during the manhattan project", "why was the manhattan project significant?", "what was the importance of communication in the manhattan project?", "why was communication important to the success of the manhattan project?", "why was the manhattan project an important achievement", "what was important about the manhattan project", "why was the manhattan project a success", "why were the scientists at the manhattan project so successful?", "why did the manhattan project really work", "what was the success of the manhattan project", "what is the importance of communication during the manhattan project", "why was the manhattan project important", "why was communication important?", "what was the importance of communication in the success of the manhattan project?", "why was the manhattan project successful?", "which statement reflects the success of the manhattan project?", "why did the manhattan project succeed", "why was the manhattan project a great success", "why was the manhattan project important"]
}
```
# Load Dataset
An example to load the dataset:
```
dataset = load_dataset('castorini/msmarco_v1_passage_doc2query-t5_expansions', data_files='d2q.jsonl.gz')
```
# Citation Information
```
@article{docTTTTTquery,
title={From doc2query to {docTTTTTquery}},
author={Nogueira, Rodrigo and Lin, Jimmy},
year={2019}
}
@article{emdt5,
author={Ronak Pradeep and Rodrigo Nogueira and Jimmy Lin},
title={The Expando-Mono-Duo Design Pattern for Text Ranking with Pretrained Sequence-to-Sequence Models},
journal={arXiv:2101.05667},
year={2021},
}
|
castorini | null | null | null | false | 322 | false | castorini/msmarco_v2_doc_doc2query-t5_expansions | 2021-11-11T17:41:32.000Z | null | false | cb336701cbfdf1de2df51de8315b27fcec566c56 | [] | [
"language:English",
"license:Apache License 2.0"
] | https://huggingface.co/datasets/castorini/msmarco_v2_doc_doc2query-t5_expansions/resolve/main/README.md | ---
language:
- English
license: "Apache License 2.0"
---
# Dataset Summary
The repo provides queries generated for the MS MARCO v2 document corpus with docTTTTTquery (sometimes written as docT5query or doc2query-T5), the latest version of the doc2query family of document expansion models. The basic idea is to train a model, that when given an input document, generates questions that the document might answer (or more broadly, queries for which the document might be relevant). These predicted questions (or queries) are then appended to the original documents, which are then indexed as before. The docTTTTTquery model gets its name from the use of T5 as the expansion model.
# Dataset Structure
All three folds (train, dev and test) share the same corpus.
An example data entry looks as follows:
```
{
'docid': '25#0',
'title': 'Autism',
'text': 'Autism is a developmental disorder characterized by difficulties with social interaction and communication, ...'
}
```
# Load Dataset
An example to load the dataset:
```
dataset = load_dataset('castorini/msmarco_v2_doc_doc2query-t5_expansions')
```
# Citation Information
```
@article{docTTTTTquery,
title={From doc2query to {docTTTTTquery}},
author={Nogueira, Rodrigo and Lin, Jimmy},
year={2019}
}
@article{emdt5,
author = "Ronak Pradeep and Rodrigo Nogueira and Jimmy Lin",
title = "The Expando-Mono-Duo Design Pattern for Text Ranking with Pretrained Sequence-to-Sequence Models",
journal = "arXiv:2101.05667",
year = 2021,
}
|
castorini | null | null | null | false | 322 | false | castorini/msmarco_v2_doc_segmented_doc2query-t5_expansions | 2021-11-02T08:13:56.000Z | null | false | 61325a80b2ff2b81642bd532483dc51d0b46a8fb | [] | [
"language:English",
"license:Apache License 2.0"
] | https://huggingface.co/datasets/castorini/msmarco_v2_doc_segmented_doc2query-t5_expansions/resolve/main/README.md | ---
language:
- English
license: "Apache License 2.0"
---
# Dataset Summary
The repo provides queries generated for the MS MARCO v2 document segmented corpus with docTTTTTquery (sometimes written as docT5query or doc2query-T5), the latest version of the doc2query family of document expansion models. The basic idea is to train a model, that when given an input document, generates questions that the document might answer (or more broadly, queries for which the document might be relevant). These predicted questions (or queries) are then appended to the original documents, which are then indexed as before. The docTTTTTquery model gets its name from the use of T5 as the expansion model.
# Dataset Structure
All three folds (train, dev and test) share the same corpus.
An example data entry looks as follows:
```
{
'docid': '25#0',
'title': 'Autism',
'text': 'Autism is a developmental disorder characterized by difficulties with social interaction and communication, ...'
}
```
# Load Dataset
An example to load the dataset:
```
dataset = load_dataset('castorini/msmarco_v2_doc_segmented_doc2query-t5_expansions', data_files='d2q/d2q.jsonl???.gz')
```
# Citation Information
```
@article{docTTTTTquery,
title={From doc2query to {docTTTTTquery}},
author={Nogueira, Rodrigo and Lin, Jimmy},
year={2019}
}
@article{emdt5,
author = "Ronak Pradeep and Rodrigo Nogueira and Jimmy Lin",
title = "The Expando-Mono-Duo Design Pattern for Text Ranking with Pretrained Sequence-to-Sequence Models",
journal = "arXiv:2101.05667",
year = 2021,
}
|
castorini | null | null | null | false | 322 | false | castorini/msmarco_v2_passage_doc2query-t5_expansions | 2021-11-02T06:37:36.000Z | null | false | 22a0c06017015ef75b33d066711b1ebc2ddb7e8e | [] | [
"language:English",
"license:Apache License 2.0"
] | https://huggingface.co/datasets/castorini/msmarco_v2_passage_doc2query-t5_expansions/resolve/main/README.md | ---
language:
- English
license: "Apache License 2.0"
---
# Dataset Summary
The repo provides queries generated for the MS MARCO v2 passage corpus with docTTTTTquery (sometimes written as docT5query or doc2query-T5), the latest version of the doc2query family of document expansion models. The basic idea is to train a model, that when given an input document, generates questions that the document might answer (or more broadly, queries for which the document might be relevant). These predicted questions (or queries) are then appended to the original documents, which are then indexed as before. The docTTTTTquery model gets its name from the use of T5 as the expansion model.
# Dataset Structure
All three folds (train, dev and test) share the same corpus. The queries are generated from this corpus.
An example data entry looks as follows:
```
{ "id": "msmarco_passage_22_0",
"predicted_queries": ["in drug combat does a zombie take more damage or die", "is the health bar the same as smash bros", "is brawlhalla health bar", "icpri league brawlhalla", "what is a battle brawlhalla", "is smash bros minecraft brawlhalla zombies", "what are the health bars on brawlhalla", "does smash bros have health bars", "is brawlhalla a health bar", "what is brawlhalla", "what is brwlhalla", "how many health bars is in brawlhalla", "is there health bar in brawlhalla", "what is boiledhalla?", "what is a good health bar in brawlhalla", "what is skills brawlhalla", "how many gobs in a brawlhalla", "is smash bros. an nsb game", "how many health bars are there in the brawlhalla", "what is brawlhalla"]
}
```
# Load Dataset
An example to load the dataset:
```
dataset = load_dataset('castorini/msmarco_v2_passage_doc2query-t5_expansions', data_files='d2q/d2q.jsonl???.gz')
```
# Citation Information
```
@article{docTTTTTquery,
title={From doc2query to {docTTTTTquery}},
author={Nogueira, Rodrigo and Lin, Jimmy},
year={2019}
}
@article{emdt5,
author={Ronak Pradeep and Rodrigo Nogueira and Jimmy Lin},
title={The Expando-Mono-Duo Design Pattern for Text Ranking with Pretrained Sequence-to-Sequence Models},
journal={arXiv:2101.05667},
year={2021},
}
|
castorini | null | null | null | false | 324 | false | castorini/nq_gar-t5_expansions | 2022-02-17T00:52:17.000Z | null | false | 30ebb5b73cf4b3a1f65de6fbd0471840dc712d34 | [] | [
"language:English",
"license:Apache License 2.0"
] | https://huggingface.co/datasets/castorini/nq_gar-t5_expansions/resolve/main/README.md | ---
language:
- English
license: "Apache License 2.0"
---
# Dataset Summary
The repo provides answer,title and sentence expansions for the Natural Questions corpus with gar-T5.
# Dataset Structure
There are dev and test folds
An example data entry of the dev split looks as follows:
```
{
"id": "1",
"predicted_answers": ["312"], "predicted_titles": ["Invisible Man"], "predicted_sentences": ["The Invisible Man First edition Author Ralph Ellison Cover artist M."]
}
```
An example data entry of the test split looks as follows:
```
{
"id": "1",
"predicted_answers": ["May 18 , 2018"], "predicted_titles": ["Deadpool 2 *** Deadpool (film) *** Deadpool 2 (soundtrack) *** X-Men in other media"], "predicted_sentences": ["Deadpool 2 was released on May 18 , 2018 , with Leitch directing from a screenplay by Rhett Reese and Paul Wernick ."]
}
```
# Load Dataset
An example to load the dataset:
```python
data_files = {"dev":"dev/dev.jsonl", "test": "test/test.jsonl"}
dataset = load_dataset('castorini/nq_gar-t5_expansions')
```
|
castorini | null | null | null | false | 323 | false | castorini/triviaqa_gar-t5_expansions | 2022-02-17T00:58:32.000Z | null | false | 5e899a9b63776d2982c72aa242cc35ecdb7073a4 | [] | [
"language:English",
"license:Apache License 2.0"
] | https://huggingface.co/datasets/castorini/triviaqa_gar-t5_expansions/resolve/main/README.md | ---
language:
- English
license: "Apache License 2.0"
---
# Dataset Summary
The repo provides answer,title and sentence expansions for the Trivia QA corpus with gar-T5.
# Dataset Structure
There are dev and test folds
An example data entry of the dev split looks as follows:
```
{
"id": "1",
"predicted_answers": ["Bz"], "predicted_titles": ["Vehicle registration plates of Belize *** Vehicle registration plate"], "predicted_sentences": ["The international code for Belize is \"\"BZ\"\"."]
}
```
An example data entry of the test split looks as follows:
```
{
"id": "1",
"predicted_answers": ["Taurus"], "predicted_titles": ["Jamie Lee Curtis *** Under the Tuscan Sun *** Angels (Jamie Lee Curtis song) *** Under the Tuscan Sun (film) *** John Michael King *** Robert Earl *** Henry Jones, Sr. *** Jamie Lee (singer) *** Under the Tuscan Sun (1974 film) *** Richard Benjamin"], "predicted_sentences": ["In July 2007, several news outlets reported that the couple had quietly married in December 2007, and that Curtis had taken a liking to one another, sharing \"\"sweet nothings\"\" about their relationship."]
}
```
# Load Dataset
An example to load the dataset:
```python
data_files = {"dev":"dev/dev.jsonl", "test": "test/test.jsonl"}
dataset = load_dataset('castorini/triviaqa_gar-t5_expansions')
```
|
ccdv | null | null | Arxiv Classification Dataset: a classification of Arxiv Papers (11 classes).
It contains 11 slightly unbalanced classes, 33k Arxiv Papers divided into 3 splits: train (23k), val (5k) and test (5k).
Copied from "Long Document Classification From Local Word Glimpses via Recurrent Attention Learning" by JUN HE LIQUN WANG LIU LIU, JIAO FENG AND HAO WU
See: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8675939
See: https://github.com/LiqunW/Long-document-dataset | false | 566 | false | ccdv/arxiv-classification | 2022-10-22T09:23:50.000Z | null | false | f9bd92144ed76200d6eb3ce73a8bd4eba9ffdc85 | [] | [
"language:en",
"task_categories:text-classification",
"tags:long context",
"task_ids:multi-class-classification",
"task_ids:topic-classification",
"size_categories:10K<n<100K"
] | https://huggingface.co/datasets/ccdv/arxiv-classification/resolve/main/README.md | ---
language: en
task_categories:
- text-classification
tags:
- long context
task_ids:
- multi-class-classification
- topic-classification
size_categories: 10K<n<100K
---
**Arxiv Classification: a classification of Arxiv Papers (11 classes).**
This dataset is intended for long context classification (documents have all > 4k tokens). \
Copied from "Long Document Classification From Local Word Glimpses via Recurrent Attention Learning"
```
@ARTICLE{8675939,
author={He, Jun and Wang, Liqun and Liu, Liu and Feng, Jiao and Wu, Hao},
journal={IEEE Access},
title={Long Document Classification From Local Word Glimpses via Recurrent Attention Learning},
year={2019},
volume={7},
number={},
pages={40707-40718},
doi={10.1109/ACCESS.2019.2907992}
}
```
* See: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8675939
* See: https://github.com/LiqunW/Long-document-dataset
It contains 11 slightly unbalanced classes, 33k Arxiv Papers divided into 3 splits: train (28k), val (2.5k) and test (2.5k).
2 configs:
* default
* no_ref, removes references to the class inside the document (eg: [cs.LG] -> [])
Compatible with [run_glue.py](https://github.com/huggingface/transformers/tree/master/examples/pytorch/text-classification) script:
```
export MODEL_NAME=roberta-base
export MAX_SEQ_LENGTH=512
python run_glue.py \
--model_name_or_path $MODEL_NAME \
--dataset_name ccdv/arxiv-classification \
--do_train \
--do_eval \
--max_seq_length $MAX_SEQ_LENGTH \
--per_device_train_batch_size 8 \
--gradient_accumulation_steps 4 \
--learning_rate 2e-5 \
--num_train_epochs 1 \
--max_eval_samples 500 \
--output_dir tmp/arxiv
``` |
ccdv | null | @inproceedings{cohan-etal-2018-discourse,
title = "A Discourse-Aware Attention Model for Abstractive Summarization of Long Documents",
author = "Cohan, Arman and
Dernoncourt, Franck and
Kim, Doo Soon and
Bui, Trung and
Kim, Seokhwan and
Chang, Walter and
Goharian, Nazli",
booktitle = "Proceedings of the 2018 Conference of the North {A}merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers)",
month = jun,
year = "2018",
address = "New Orleans, Louisiana",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/N18-2097",
doi = "10.18653/v1/N18-2097",
pages = "615--621",
abstract = "Neural abstractive summarization models have led to promising results in summarizing relatively short documents. We propose the first model for abstractive summarization of single, longer-form documents (e.g., research papers). Our approach consists of a new hierarchical encoder that models the discourse structure of a document, and an attentive discourse-aware decoder to generate the summary. Empirical results on two large-scale datasets of scientific papers show that our model significantly outperforms state-of-the-art models.",
} | Arxiv dataset for summarization.
From paper: A Discourse-Aware Attention Model for Abstractive Summarization of Long Documents" by A. Cohan et al.
See: https://aclanthology.org/N18-2097.pdf
See: https://github.com/armancohan/long-summarization | false | 20,803 | false | ccdv/arxiv-summarization | 2022-10-24T20:31:40.000Z | null | false | d6b009c8dea00d6db75a680595b4340546ae4020 | [] | [
"language:en",
"multilinguality:monolingual",
"size_categories:100K<n<1M",
"task_categories:summarization",
"task_categories:text-generation",
"tags:conditional-text-generation"
] | https://huggingface.co/datasets/ccdv/arxiv-summarization/resolve/main/README.md | ---
language:
- en
multilinguality:
- monolingual
size_categories:
- 100K<n<1M
task_categories:
- summarization
- text-generation
task_ids: []
tags:
- conditional-text-generation
---
# Arxiv dataset for summarization
Dataset for summarization of long documents.\
Adapted from this [repo](https://github.com/armancohan/long-summarization).\
Note that original data are pre-tokenized so this dataset returns " ".join(text) and add "\n" for paragraphs. \
This dataset is compatible with the [`run_summarization.py`](https://github.com/huggingface/transformers/tree/master/examples/pytorch/summarization) script from Transformers if you add this line to the `summarization_name_mapping` variable:
```python
"ccdv/arxiv-summarization": ("article", "abstract")
```
### Data Fields
- `id`: paper id
- `article`: a string containing the body of the paper
- `abstract`: a string containing the abstract of the paper
### Data Splits
This dataset has 3 splits: _train_, _validation_, and _test_. \
Token counts are white space based.
| Dataset Split | Number of Instances | Avg. tokens |
| ------------- | --------------------|:----------------------|
| Train | 203,037 | 6038 / 299 |
| Validation | 6,436 | 5894 / 172 |
| Test | 6,440 | 5905 / 174 |
# Cite original article
```
@inproceedings{cohan-etal-2018-discourse,
title = "A Discourse-Aware Attention Model for Abstractive Summarization of Long Documents",
author = "Cohan, Arman and
Dernoncourt, Franck and
Kim, Doo Soon and
Bui, Trung and
Kim, Seokhwan and
Chang, Walter and
Goharian, Nazli",
booktitle = "Proceedings of the 2018 Conference of the North {A}merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers)",
month = jun,
year = "2018",
address = "New Orleans, Louisiana",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/N18-2097",
doi = "10.18653/v1/N18-2097",
pages = "615--621",
abstract = "Neural abstractive summarization models have led to promising results in summarizing relatively short documents. We propose the first model for abstractive summarization of single, longer-form documents (e.g., research papers). Our approach consists of a new hierarchical encoder that models the discourse structure of a document, and an attentive discourse-aware decoder to generate the summary. Empirical results on two large-scale datasets of scientific papers show that our model significantly outperforms state-of-the-art models.",
}
```
|
ccdv | null | @article{DBLP:journals/corr/SeeLM17,
author = {Abigail See and
Peter J. Liu and
Christopher D. Manning},
title = {Get To The Point: Summarization with Pointer-Generator Networks},
journal = {CoRR},
volume = {abs/1704.04368},
year = {2017},
url = {http://arxiv.org/abs/1704.04368},
archivePrefix = {arXiv},
eprint = {1704.04368},
timestamp = {Mon, 13 Aug 2018 16:46:08 +0200},
biburl = {https://dblp.org/rec/bib/journals/corr/SeeLM17},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
@inproceedings{hermann2015teaching,
title={Teaching machines to read and comprehend},
author={Hermann, Karl Moritz and Kocisky, Tomas and Grefenstette, Edward and Espeholt, Lasse and Kay, Will and Suleyman, Mustafa and Blunsom, Phil},
booktitle={Advances in neural information processing systems},
pages={1693--1701},
year={2015}
} | CNN/DailyMail non-anonymized summarization dataset.
There are two features:
- article: text of news article, used as the document to be summarized
- highlights: joined text of highlights with <s> and </s> around each
highlight, which is the target summary | false | 3,691 | false | ccdv/cnn_dailymail | 2022-10-24T20:31:59.000Z | cnn-daily-mail-1 | false | dc2ce3bd19d8e323365bc1a244f3dd32e02d4f22 | [] | [
"annotations_creators:no-annotation",
"language_creators:found",
"language:en",
"license:apache-2.0",
"multilinguality:monolingual",
"size_categories:100K<n<1M",
"source_datasets:original",
"task_categories:summarization",
"task_categories:text-generation",
"tags:conditional-text-generation"
] | https://huggingface.co/datasets/ccdv/cnn_dailymail/resolve/main/README.md | ---
annotations_creators:
- no-annotation
language_creators:
- found
language:
- en
license:
- apache-2.0
multilinguality:
- monolingual
size_categories:
- 100K<n<1M
source_datasets:
- original
task_categories:
- summarization
- text-generation
task_ids: []
paperswithcode_id: cnn-daily-mail-1
pretty_name: CNN / Daily Mail
tags:
- conditional-text-generation
---
**Copy of the [cnn_dailymail](https://huggingface.co/datasets/cnn_dailymail) dataset fixing the "NotADirectoryError: [Errno 20]".**
# Dataset Card for CNN Dailymail Dataset
## 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:** [CNN / DailyMail Dataset repository](https://github.com/abisee/cnn-dailymail)
- **Paper:** [Abstractive Text Summarization Using Sequence-to-Sequence RNNs and Beyond](https://papers.nips.cc/paper/5945-teaching-machines-to-read-and-comprehend.pdf), [Get To The Point: Summarization with Pointer-Generator Networks](https://www.aclweb.org/anthology/K16-1028.pdf)
- **Leaderboard:** [Papers with Code leaderboard for CNN / Dailymail Dataset](https://paperswithcode.com/sota/document-summarization-on-cnn-daily-mail)
- **Point of Contact:** [Abigail See](mailto:abisee@stanford.edu)
### Dataset Summary
The CNN / DailyMail Dataset is an English-language dataset containing just over 300k unique news articles as written by journalists at CNN and the Daily Mail. The current version supports both extractive and abstractive summarization, though the original version was created for machine reading and comprehension and abstractive question answering.
### Supported Tasks and Leaderboards
- 'summarization': [Versions 2.0.0 and 3.0.0 of the CNN / DailyMail Dataset](https://www.aclweb.org/anthology/K16-1028.pdf) can be used to train a model for abstractive and extractive summarization ([Version 1.0.0](https://papers.nips.cc/paper/5945-teaching-machines-to-read-and-comprehend.pdf) was developed for machine reading and comprehension and abstractive question answering). The model performance is measured by how high the output summary's [ROUGE](https://huggingface.co/metrics/rouge) score for a given article is when compared to the highlight as written by the original article author. [Zhong et al (2020)](https://www.aclweb.org/anthology/2020.acl-main.552.pdf) report a ROUGE-1 score of 44.41 when testing a model trained for extractive summarization. See the [Papers With Code leaderboard](https://paperswithcode.com/sota/document-summarization-on-cnn-daily-mail) for more models.
### Languages
The BCP-47 code for English as generally spoken in the United States is en-US and the BCP-47 code for English as generally spoken in the United Kingdom is en-GB. It is unknown if other varieties of English are represented in the data.
## Dataset Structure
### Data Instances
For each instance, there is a string for the article, a string for the highlights, and a string for the id. See the [CNN / Daily Mail dataset viewer](https://huggingface.co/datasets/viewer/?dataset=cnn_dailymail&config=3.0.0) to explore more examples.
```
{'id': '0054d6d30dbcad772e20b22771153a2a9cbeaf62',
'article': '(CNN) -- An American woman died aboard a cruise ship that docked at Rio de Janeiro on Tuesday, the same ship on which 86 passengers previously fell ill, according to the state-run Brazilian news agency, Agencia Brasil. The American tourist died aboard the MS Veendam, owned by cruise operator Holland America. Federal Police told Agencia Brasil that forensic doctors were investigating her death. The ship's doctors told police that the woman was elderly and suffered from diabetes and hypertension, according the agency. The other passengers came down with diarrhea prior to her death during an earlier part of the trip, the ship's doctors said. The Veendam left New York 36 days ago for a South America tour.'
'highlights': 'The elderly woman suffered from diabetes and hypertension, ship's doctors say .\nPreviously, 86 passengers had fallen ill on the ship, Agencia Brasil says .'}
```
The average token count for the articles and the highlights are provided below:
| Feature | Mean Token Count |
| ---------- | ---------------- |
| Article | 781 |
| Highlights | 56 |
### Data Fields
- `id`: a string containing the heximal formated SHA1 hash of the url where the story was retrieved from
- `article`: a string containing the body of the news article
- `highlights`: a string containing the highlight of the article as written by the article author
### Data Splits
The CNN/DailyMail dataset has 3 splits: _train_, _validation_, and _test_. Below are the statistics for Version 3.0.0 of the dataset.
| Dataset Split | Number of Instances in Split |
| ------------- | ------------------------------------------- |
| Train | 287,113 |
| Validation | 13,368 |
| Test | 11,490 |
## Dataset Creation
### Curation Rationale
Version 1.0.0 aimed to support supervised neural methodologies for machine reading and question answering with a large amount of real natural language training data and released about 313k unique articles and nearly 1M Cloze style questions to go with the articles. Versions 2.0.0 and 3.0.0 changed the structure of the dataset to support summarization rather than question answering. Version 3.0.0 provided a non-anonymized version of the data, whereas both the previous versions were preprocessed to replace named entities with unique identifier labels.
### Source Data
#### Initial Data Collection and Normalization
The data consists of news articles and highlight sentences. In the question answering setting of the data, the articles are used as the context and entities are hidden one at a time in the highlight sentences, producing Cloze style questions where the goal of the model is to correctly guess which entity in the context has been hidden in the highlight. In the summarization setting, the highlight sentences are concatenated to form a summary of the article. The CNN articles were written between April 2007 and April 2015. The Daily Mail articles were written between June 2010 and April 2015.
The code for the original data collection is available at <https://github.com/deepmind/rc-data>. The articles were downloaded using archives of <www.cnn.com> and <www.dailymail.co.uk> on the Wayback Machine. Articles were not included in the Version 1.0.0 collection if they exceeded 2000 tokens. Due to accessibility issues with the Wayback Machine, Kyunghyun Cho has made the datasets available at <https://cs.nyu.edu/~kcho/DMQA/>. An updated version of the code that does not anonymize the data is available at <https://github.com/abisee/cnn-dailymail>.
Hermann et al provided their own tokenization script. The script provided by See uses the PTBTokenizer. It also lowercases the text and adds periods to lines missing them.
#### Who are the source language producers?
The text was written by journalists at CNN and the Daily Mail.
### Annotations
The dataset does not contain any additional annotations.
#### Annotation process
[N/A]
#### Who are the annotators?
[N/A]
### Personal and Sensitive Information
Version 3.0 is not anonymized, so individuals' names can be found in the dataset. Information about the original author is not included in the dataset.
## Considerations for Using the Data
### Social Impact of Dataset
The purpose of this dataset is to help develop models that can summarize long paragraphs of text in one or two sentences.
This task is useful for efficiently presenting information given a large quantity of text. It should be made clear that any summarizations produced by models trained on this dataset are reflective of the language used in the articles, but are in fact automatically generated.
### Discussion of Biases
[Bordia and Bowman (2019)](https://www.aclweb.org/anthology/N19-3002.pdf) explore measuring gender bias and debiasing techniques in the CNN / Dailymail dataset, the Penn Treebank, and WikiText-2. They find the CNN / Dailymail dataset to have a slightly lower gender bias based on their metric compared to the other datasets, but still show evidence of gender bias when looking at words such as 'fragile'.
Because the articles were written by and for people in the US and the UK, they will likely present specifically US and UK perspectives and feature events that are considered relevant to those populations during the time that the articles were published.
### Other Known Limitations
News articles have been shown to conform to writing conventions in which important information is primarily presented in the first third of the article [(Kryściński et al, 2019)](https://www.aclweb.org/anthology/D19-1051.pdf). [Chen et al (2016)](https://www.aclweb.org/anthology/P16-1223.pdf) conducted a manual study of 100 random instances of the first version of the dataset and found 25% of the samples to be difficult even for humans to answer correctly due to ambiguity and coreference errors.
It should also be noted that machine-generated summarizations, even when extractive, may differ in truth values when compared to the original articles.
## Additional Information
### Dataset Curators
The data was originally collected by Karl Moritz Hermann, Tomáš Kočiský, Edward Grefenstette, Lasse Espeholt, Will Kay, Mustafa Suleyman, and Phil Blunsom of Google DeepMind. Tomáš Kočiský and Phil Blunsom are also affiliated with the University of Oxford. They released scripts to collect and process the data into the question answering format.
Ramesh Nallapati, Bowen Zhou, Cicero dos Santos, and Bing Xiang of IMB Watson and Çağlar Gu̇lçehre of Université de Montréal modified Hermann et al's collection scripts to restore the data to a summary format. They also produced both anonymized and non-anonymized versions.
The code for the non-anonymized version is made publicly available by Abigail See of Stanford University, Peter J. Liu of Google Brain and Christopher D. Manning of Stanford University at <https://github.com/abisee/cnn-dailymail>. The work at Stanford University was supported by the DARPA DEFT ProgramAFRL contract no. FA8750-13-2-0040.
### Licensing Information
The CNN / Daily Mail dataset version 1.0.0 is released under the [Apache-2.0 License](http://www.apache.org/licenses/LICENSE-2.0).
### Citation Information
```
@inproceedings{see-etal-2017-get,
title = "Get To The Point: Summarization with Pointer-Generator Networks",
author = "See, Abigail and
Liu, Peter J. and
Manning, Christopher D.",
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://www.aclweb.org/anthology/P17-1099",
doi = "10.18653/v1/P17-1099",
pages = "1073--1083",
abstract = "Neural sequence-to-sequence models have provided a viable new approach for abstractive text summarization (meaning they are not restricted to simply selecting and rearranging passages from the original text). However, these models have two shortcomings: they are liable to reproduce factual details inaccurately, and they tend to repeat themselves. In this work we propose a novel architecture that augments the standard sequence-to-sequence attentional model in two orthogonal ways. First, we use a hybrid pointer-generator network that can copy words from the source text via pointing, which aids accurate reproduction of information, while retaining the ability to produce novel words through the generator. Second, we use coverage to keep track of what has been summarized, which discourages repetition. We apply our model to the CNN / Daily Mail summarization task, outperforming the current abstractive state-of-the-art by at least 2 ROUGE points.",
}
```
```
@inproceedings{DBLP:conf/nips/HermannKGEKSB15,
author={Karl Moritz Hermann and Tomás Kociský and Edward Grefenstette and Lasse Espeholt and Will Kay and Mustafa Suleyman and Phil Blunsom},
title={Teaching Machines to Read and Comprehend},
year={2015},
cdate={1420070400000},
pages={1693-1701},
url={http://papers.nips.cc/paper/5945-teaching-machines-to-read-and-comprehend},
booktitle={NIPS},
crossref={conf/nips/2015}
}
```
### Contributions
Thanks to [@thomwolf](https://github.com/thomwolf), [@lewtun](https://github.com/lewtun), [@jplu](https://github.com/jplu), [@jbragg](https://github.com/jbragg), [@patrickvonplaten](https://github.com/patrickvonplaten) and [@mcmillanmajora](https://github.com/mcmillanmajora) for adding this dataset.
|
ccdv | null | @misc{huang2021efficient,
title={Efficient Attentions for Long Document Summarization},
author={Luyang Huang and Shuyang Cao and Nikolaus Parulian and Heng Ji and Lu Wang},
year={2021},
eprint={2104.02112},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
} | GovReport dataset for summarization.
From paper: Efficient Attentions for Long Document Summarization" by L. Huang et al.
See: https://arxiv.org/pdf/2104.02112.pdf
See: https://github.com/luyang-huang96/LongDocSum | false | 453 | false | ccdv/govreport-summarization | 2022-10-24T20:32:47.000Z | null | false | b949637ab41c9f668a4b83cea46c80b489c02290 | [] | [
"arxiv:2104.02112",
"language:en",
"multilinguality:monolingual",
"size_categories:10K<n<100K",
"task_categories:summarization",
"task_categories:text-generation",
"tags:conditional-text-generation"
] | https://huggingface.co/datasets/ccdv/govreport-summarization/resolve/main/README.md | ---
language:
- en
multilinguality:
- monolingual
size_categories:
- 10K<n<100K
task_categories:
- summarization
- text-generation
task_ids: []
tags:
- conditional-text-generation
---
# GovReport dataset for summarization
Dataset for summarization of long documents.\
Adapted from this [repo](https://github.com/luyang-huang96/LongDocSum) and this [paper](https://arxiv.org/pdf/2104.02112.pdf)\
This dataset is compatible with the [`run_summarization.py`](https://github.com/huggingface/transformers/tree/master/examples/pytorch/summarization) script from Transformers if you add this line to the `summarization_name_mapping` variable:
```python
"ccdv/govreport-summarization": ("report", "summary")
```
### Data Fields
- `id`: paper id
- `report`: a string containing the body of the report
- `summary`: a string containing the summary of the report
### Data Splits
This dataset has 3 splits: _train_, _validation_, and _test_. \
Token counts with a RoBERTa tokenizer.
| Dataset Split | Number of Instances | Avg. tokens |
| ------------- | --------------------|:----------------------|
| Train | 17,517 | < 9,000 / < 500 |
| Validation | 973 | < 9,000 / < 500 |
| Test | 973 | < 9,000 / < 500 |
# Cite original article
```
@misc{huang2021efficient,
title={Efficient Attentions for Long Document Summarization},
author={Luyang Huang and Shuyang Cao and Nikolaus Parulian and Heng Ji and Lu Wang},
year={2021},
eprint={2104.02112},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
```
|
ccdv | null | null | Patent Classification Dataset: a classification of Patents (9 classes).
It contains 9 unbalanced classes, 35k Patents and summaries divided into 3 splits: train (25k), val (5k) and test (5k).
Data are sampled from "BIGPATENT: A Large-Scale Dataset for Abstractive and Coherent Summarization." by Eva Sharma, Chen Li and Lu Wang
See: https://aclanthology.org/P19-1212.pdf
See: https://evasharma.github.io/bigpatent/ | false | 480 | false | ccdv/patent-classification | 2022-10-22T09:25:36.000Z | null | false | 2f38a1dfdecfacee0184d74eaeafd3c0fb49d2a6 | [] | [
"language:en",
"task_categories:text-classification",
"tags:long context",
"task_ids:multi-class-classification",
"task_ids:topic-classification",
"size_categories:10K<n<100K"
] | https://huggingface.co/datasets/ccdv/patent-classification/resolve/main/README.md | ---
language: en
task_categories:
- text-classification
tags:
- long context
task_ids:
- multi-class-classification
- topic-classification
size_categories: 10K<n<100K
---
**Patent Classification: a classification of Patents and abstracts (9 classes).**
This dataset is intended for long context classification (non abstract documents are longer that 512 tokens). \
Data are sampled from "BIGPATENT: A Large-Scale Dataset for Abstractive and Coherent Summarization." by Eva Sharma, Chen Li and Lu Wang
* See: https://aclanthology.org/P19-1212.pdf
* See: https://evasharma.github.io/bigpatent/
It contains 9 unbalanced classes, 35k Patents and abstracts divided into 3 splits: train (25k), val (5k) and test (5k).
**Note that documents are uncased and space separated (by authors)**
Compatible with [run_glue.py](https://github.com/huggingface/transformers/tree/master/examples/pytorch/text-classification) script:
```
export MODEL_NAME=roberta-base
export MAX_SEQ_LENGTH=512
python run_glue.py \
--model_name_or_path $MODEL_NAME \
--dataset_name ccdv/patent-classification \
--do_train \
--do_eval \
--max_seq_length $MAX_SEQ_LENGTH \
--per_device_train_batch_size 8 \
--gradient_accumulation_steps 4 \
--learning_rate 2e-5 \
--num_train_epochs 1 \
--max_eval_samples 500 \
--output_dir tmp/patent
``` |
ccdv | null | @inproceedings{cohan-etal-2018-discourse,
title = "A Discourse-Aware Attention Model for Abstractive Summarization of Long Documents",
author = "Cohan, Arman and
Dernoncourt, Franck and
Kim, Doo Soon and
Bui, Trung and
Kim, Seokhwan and
Chang, Walter and
Goharian, Nazli",
booktitle = "Proceedings of the 2018 Conference of the North {A}merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers)",
month = jun,
year = "2018",
address = "New Orleans, Louisiana",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/N18-2097",
doi = "10.18653/v1/N18-2097",
pages = "615--621",
abstract = "Neural abstractive summarization models have led to promising results in summarizing relatively short documents. We propose the first model for abstractive summarization of single, longer-form documents (e.g., research papers). Our approach consists of a new hierarchical encoder that models the discourse structure of a document, and an attentive discourse-aware decoder to generate the summary. Empirical results on two large-scale datasets of scientific papers show that our model significantly outperforms state-of-the-art models.",
} | PubMed dataset for summarization.
From paper: A Discourse-Aware Attention Model for Abstractive Summarization of Long Documents" by A. Cohan et al.
See: https://aclanthology.org/N18-2097.pdf
See: https://github.com/armancohan/long-summarization | false | 7,727 | false | ccdv/pubmed-summarization | 2022-10-24T20:33:04.000Z | null | false | 26155ccf2b18393a38a05fafc26c66a068974839 | [] | [
"language:en",
"multilinguality:monolingual",
"size_categories:100K<n<1M",
"task_categories:summarization",
"task_categories:text-generation",
"tags:conditional-text-generation"
] | https://huggingface.co/datasets/ccdv/pubmed-summarization/resolve/main/README.md | ---
language:
- en
multilinguality:
- monolingual
size_categories:
- 100K<n<1M
task_categories:
- summarization
- text-generation
task_ids: []
tags:
- conditional-text-generation
---
# PubMed dataset for summarization
Dataset for summarization of long documents.\
Adapted from this [repo](https://github.com/armancohan/long-summarization).\
Note that original data are pre-tokenized so this dataset returns " ".join(text) and add "\n" for paragraphs. \
This dataset is compatible with the [`run_summarization.py`](https://github.com/huggingface/transformers/tree/master/examples/pytorch/summarization) script from Transformers if you add this line to the `summarization_name_mapping` variable:
```python
"ccdv/pubmed-summarization": ("article", "abstract")
```
### Data Fields
- `id`: paper id
- `article`: a string containing the body of the paper
- `abstract`: a string containing the abstract of the paper
### Data Splits
This dataset has 3 splits: _train_, _validation_, and _test_. \
Token counts are white space based.
| Dataset Split | Number of Instances | Avg. tokens |
| ------------- | --------------------|:----------------------|
| Train | 119,924 | 3043 / 215 |
| Validation | 6,633 | 3111 / 216 |
| Test | 6,658 | 3092 / 219 |
# Cite original article
```
@inproceedings{cohan-etal-2018-discourse,
title = "A Discourse-Aware Attention Model for Abstractive Summarization of Long Documents",
author = "Cohan, Arman and
Dernoncourt, Franck and
Kim, Doo Soon and
Bui, Trung and
Kim, Seokhwan and
Chang, Walter and
Goharian, Nazli",
booktitle = "Proceedings of the 2018 Conference of the North {A}merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers)",
month = jun,
year = "2018",
address = "New Orleans, Louisiana",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/N18-2097",
doi = "10.18653/v1/N18-2097",
pages = "615--621",
abstract = "Neural abstractive summarization models have led to promising results in summarizing relatively short documents. We propose the first model for abstractive summarization of single, longer-form documents (e.g., research papers). Our approach consists of a new hierarchical encoder that models the discourse structure of a document, and an attentive discourse-aware decoder to generate the summary. Empirical results on two large-scale datasets of scientific papers show that our model significantly outperforms state-of-the-art models.",
}
```
|
cdleong | null | \\r\n@InProceedings{huggingface:dataset,
title = {A great new dataset},
author={huggingface, Inc.
},
year={2020}
} | \\r\nPig-latin machine and English parallel machine translation corpus.
Based on
The Project Gutenberg EBook of "De Bello Gallico" and Other Commentaries
https://www.gutenberg.org/ebooks/10657
Converted to pig-latin with https://github.com/bpabel/piglatin | false | 323 | false | cdleong/piglatin-mt | 2022-10-24T19:22:09.000Z | null | false | 464088ad69bd568eba869f3af6bc2f16a9cd9a5c | [] | [
"language:en",
"license:mit",
"multilinguality:translation",
"size_categories:10K<n<100K",
"source_datasets:original",
"task_categories:translation",
"language_details:eng and engyay"
] | https://huggingface.co/datasets/cdleong/piglatin-mt/resolve/main/README.md | ---
language:
- en
license:
- mit
multilinguality:
- translation
size_categories:
- 10K<n<100K
source_datasets:
- original
task_categories:
- translation
task_ids: []
language_details: eng and engyay
---
## Dataset Description
- **Homepage:** cdleong.github.io
# Dataset Summary:
Pig-latin machine and English parallel machine translation corpus.
Based on [The Project Gutenberg EBook of "De Bello Gallico" and Other Commentaries](https://www.gutenberg.org/ebooks/10657)
Converted to pig-latin with https://github.com/bpabel/piglatin
Blank lines removed.
## Dataset Structure
```
DatasetDict({
train: Dataset({
features: ['translation'],
num_rows: 14778
})
validation: Dataset({
features: ['translation'],
num_rows: 1000
})
})
```
### Data Instances
```
{
'translation':
{
'eng': 'thrown into disorder they returned with more precipitation than is usual',
'engyay': 'own-thray into-ay isorder-day ey-thay eturned-ray ith-way ore-may ecipitation-pray an-thay is-ay usual-ay'
}
}
```
### Data Fields
- `translation`: a dictionary containing two strings paired with a key indicating the corresponding language.
### Data Splits
- `train`: most of the data, 13,232 samples total.
- `dev`: 1k holdout samples, created with the datasets.train_test_split() function |
cdleong | null | @InProceedings{huggingface:dataset,
title = {A great new dataset},
author={huggingface, Inc.
},
year={2020}
} | This new dataset is designed to solve this great NLP task and is crafted with a lot of care. | false | 167 | false | cdleong/temp_africaNLP_keyword_spotting_for_african_languages | 2022-10-25T09:07:32.000Z | null | false | edb6563c1ba616922132466f1a969807bba8651e | [] | [
"language:wo",
"language:fuc",
"language:srr",
"language:mnk",
"language:snk"
] | https://huggingface.co/datasets/cdleong/temp_africaNLP_keyword_spotting_for_african_languages/resolve/main/README.md | ---
language:
- wo
- fuc
- srr
- mnk
- snk
---
## Dataset Description
- **Homepage:** https://zenodo.org/record/4661645
TEMPORARY TEST DATASET
Not for actual use! Attempting to test out a dataset script for loading https://zenodo.org/record/4661645
|
cemigo | null | null | null | false | 165 | false | cemigo/taylor_vs_shakes | 2021-03-14T23:45:59.000Z | null | false | 2784446f8e97c8a4a2ce7242bb8a7537b36ff3dc | [] | [] | https://huggingface.co/datasets/cemigo/taylor_vs_shakes/resolve/main/README.md | This dataset has 336 pieces of quotes from William Shakespeare and Taylor Swift (labeled) for supervised classification.
Source: https://www.kaggle.com/kellylougheed/tswift-vs-shakespeare |
cfilt | null | null | null | false | 499 | false | cfilt/iitb-english-hindi | 2022-04-26T13:50:22.000Z | null | false | 445aaa1baafa9bf671df3cffaeb149ec44410461 | [] | [] | https://huggingface.co/datasets/cfilt/iitb-english-hindi/resolve/main/README.md | <p align="center"><img src="https://huggingface.co/datasets/cfilt/HiNER-collapsed/raw/main/cfilt-dark-vec.png" alt="Computation for Indian Language Technology Logo" width="150" height="150"/></p>
# IITB-English-Hindi Parallel Corpus
[](https://creativecommons.org/licenses/by-nc/4.0/)
[](https://twitter.com/cfiltnlp)
[](https://twitter.com/PeopleCentredAI)
## About
The IIT Bombay English-Hindi corpus contains parallel corpus for English-Hindi as well as monolingual Hindi corpus collected from a variety of existing sources and corpora developed at the Center for Indian Language Technology, IIT Bombay over the years. This page describes the corpus. This corpus has been used at the Workshop on Asian Language Translation Shared Task since 2016 the Hindi-to-English and English-to-Hindi languages pairs and as a pivot language pair for the Hindi-to-Japanese and Japanese-to-Hindi language pairs.
The complete details of this corpus are available at [this URL](https://www.cfilt.iitb.ac.in/iitb_parallel/). We also provide this parallel corpus via browser download from the same URL. We also provide a monolingual Hindi corpus on the same URL.
### Recent Updates
* Version 3.1 - December 2021 - Added 49,400 sentence pairs to the parallel corpus.
* Version 3.0 - August 2020 - Added ~47,000 sentence pairs to the parallel corpus.
## Usage
We provide a notebook that shows how to import the IITB English-Hindi Parallel Corpus from the HuggingFace datasets repository. The notebook also shows how to segment the corpus using BPE tokenization which can be used to train an English-Hindi MT System.
[https://github.com/cfiltnlp/IITB-English-Hindi-PC](https://github.com/cfiltnlp/IITB-English-Hindi-PC)
## Other
You can find a catalogue of other English-Hindi and other Indian language parallel corpora here: [Indic NLP Catalog](https://github.com/indicnlpweb/indicnlp_catalog)
## Maintainer(s)
[Diptesh Kanojia](https://dipteshkanojia.github.io)<br/>
Shivam Mhasker<br/>
## Citation
If you use this corpus or its derivate resources for your research, kindly cite it as follows:
Anoop Kunchukuttan, Pratik Mehta, Pushpak Bhattacharyya. The IIT Bombay English-Hindi Parallel Corpus. Language Resources and Evaluation Conference. 2018.
### BiBTeX Citation
```latex
@inproceedings{kunchukuttan-etal-2018-iit,
title = "The {IIT} {B}ombay {E}nglish-{H}indi Parallel Corpus",
author = "Kunchukuttan, Anoop and
Mehta, Pratik and
Bhattacharyya, Pushpak",
booktitle = "Proceedings of the Eleventh International Conference on Language Resources and Evaluation ({LREC} 2018)",
month = may,
year = "2018",
address = "Miyazaki, Japan",
publisher = "European Language Resources Association (ELRA)",
url = "https://aclanthology.org/L18-1548",
}
```
|
cgarciae | null | # @article{lecun2010mnist,
# title={MNIST handwritten digit database},
# author={LeCun, Yann and Cortes, Corinna and Burges, CJ},
# journal={ATT Labs [Online]. Available: http://yann.lecun.com/exdb/mnist},
# volume={2},
# year={2010}
# }
# | The MNIST dataset consists of 70,000 28x28 black-and-white points in 10 classes (one for each digits), with 7,000
points per class. There are 60,000 training points and 10,000 test points. | false | 326 | false | cgarciae/point-cloud-mnist | 2021-10-31T23:09:55.000Z | null | false | b88be4d36f97e51173120d42cd35ce2ffa074cc9 | [] | [] | https://huggingface.co/datasets/cgarciae/point-cloud-mnist/resolve/main/README.md | # Point CLoud MNIST
A point cloud version of the original MNIST.
## Getting Started
```python
import matplotlib.pyplot as plt
import numpy as np
from datasets import load_dataset
# load dataset
dataset = load_dataset("cgarciae/point-cloud-mnist")
dataset.set_format("np")
# get numpy arrays
X_train = dataset["train"]["points"]
y_train = dataset["train"]["label"]
X_test = dataset["test"]["points"]
y_test = dataset["test"]["label"]
# plot some training samples
figure = plt.figure(figsize=(10, 10))
for i in range(3):
for j in range(3):
k = 3 * i + j
plt.subplot(3, 3, k + 1)
idx = np.random.randint(0, len(X_train))
plt.title(f"{y_train[idx]}")
plt.scatter(X_train[idx, :, 0], X_train[idx, :, 1])
plt.show()
```
## Format
* `points`: `(batch, point, 3)` array of uint8.
* `label`: `(batch, 1)` array of uint8.
Where `point` is the number of points in the point cloud. Points have no order and were shuffled when creating the data. Each point has the structure `[x, y, v]` where:
* `x`: is the x coordinate of the point in the image.
* `y`: is the y coordinate of the point in the image.
* `v`: is the value of the pixel at the point in the image.
Samples are padded with `0`s such that `point = 351` since its the largest number of non-zero pixels per image in the original dataset. You can tell apart padding point because they are the only ones where `v = 0`.
Here is the distribution of non-zero pixels in the MNIST:
 |
chau | null | null | null | false | 324 | false | chau/ink_test01 | 2022-02-15T09:15:56.000Z | null | false | 8b8191c92578f5f381bd7020eddbb7c334d414eb | [] | [
"license:other"
] | https://huggingface.co/datasets/chau/ink_test01/resolve/main/README.md | ---
license: other
---
|
chenghao | null | null | null | false | 324 | false | chenghao/scielo_books | 2022-07-01T18:34:59.000Z | null | false | c3d46ee0b1969347cb803449156be9a59e275ae7 | [] | [
"annotations_creators:no-annotation",
"language_creators:found",
"language:en",
"language:pt",
"language:es",
"license:cc-by-nc-sa-3.0",
"multilinguality:multilingual",
"size_categories:n<1K",
"source_datasets:original",
"task_ids:language-modeling"
] | https://huggingface.co/datasets/chenghao/scielo_books/resolve/main/README.md | ---
annotations_creators:
- no-annotation
language_creators:
- found
language:
- en
- pt
- es
license:
- cc-by-nc-sa-3.0
multilinguality:
- multilingual
paperswithcode_id: null
size_categories:
- n<1K
source_datasets:
- original
task_categories:
- sequence-modeling
task_ids:
- language-modeling
---
## Dataset Description
- **Homepage:** [scielo.org](https://search.livros.scielo.org/search/?fb=&where=BOOK&filter%5Bis_comercial_filter%5D%5B%5D=f)
### Dataset Summary
This dataset contains all text from open-access PDFs on [scielo.org](https://search.livros.scielo.org/search/?fb=&where=BOOK&filter%5Bis_comercial_filter%5D%5B%5D=f). As of Dec. 5 2021, the total number of books available is 962. Some of them are not in native PDF format (e.g. scanned images) though.
### Supported Tasks and Leaderboards
- `sequence-modeling` or `language-modeling`: The dataset can be used to train a language model.
### Languages
As of Dec. 5 2021, there are 902 books in Portuguese, 55 in Spanish, and 5 in English.
## Dataset Structure
### Data Instances
Provide an JSON-formatted example and brief description of a typical instance in the dataset. If available, provide a link to further examples.
```
{
"sbid":"23pcw",
"id":"23pcw",
"shortname":"",
"title":"Educa\u00e7\u00e3o, sa\u00fade e esporte: novos\tdesafios \u00e0 Educa\u00e7\u00e3o F\u00edsica",
"eisbn":"9788574554907",
"isbn":"9788574554273",
"author":"Farias, Gelcemar Oliveira; Nascimento, Juarez Vieira do",
"corporate_authors":"",
"translators":"",
"coordinators":"",
"editors":"",
"others":"",
"organizers":"",
"collaborators":"",
"publisher":"Editus",
"language":"pt",
"year": 2016,
"synopsis":"\"A colet\u00e2nea contempla cap\u00edtulos que discutem a Educa\u00e7\u00e3o F\u00edsica a partir dos pressupostos da Educa\u00e7\u00e3o, da Sa\u00fade e do Esporte, enquanto importante desafio do momento atual e diante dos avan\u00e7os e das mudan\u00e7as que se consolidaram na forma\u00e7\u00e3o inicial em Educa\u00e7\u00e3o F\u00edsica. A obra convida a todos para a realiza\u00e7\u00e3o de futuras investiga\u00e7\u00f5es, no sentido de concentrar esfor\u00e7os para o fortalecimento de n\u00facleos de estudos e a sistematiza\u00e7\u00e3o de linhas de pesquisa.\"",
"format":"",
"type":"book",
"is_public":"true",
"is_comercial":"false",
"publication_date":"2018-11-07",
"_version_":"1718206093473087488",
"pdf_url":"http://books.scielo.org//id/23pcw/pdf/farias-9788574554907.pdf",
"pdf_filename":"farias-9788574554907.pdf",
"metadata_filename":"farias-9788574554907.json",
"text":"..."
}
```
### Data Fields
All fields are of string type except `year`.
### Data Splits
All records are in the default `train` split.
## Dataset Creation
### Curation Rationale
Part of the big science efforts to create lanague modeling datasets.
### Source Data
[scielo.org](https://search.livros.scielo.org/search/?fb=&where=BOOK&filter%5Bis_comercial_filter%5D%5B%5D=f)
#### Initial Data Collection and Normalization
All PDFs are directly downloaded from the website and text is extracted with [pdftotext](https://pypi.org/project/pdftotext/) lib.
#### Who are the source language producers?
NA
### Annotations
No annotation is available.
#### Annotation process
NA
#### Who are the annotators?
NA
### Personal and Sensitive Information
NA
## Considerations for Using the Data
### Social Impact of Dataset
NA
### Discussion of Biases
NA
### Other Known Limitations
NA
## Additional Information
### Dataset Curators
[@chenghao](https://huggingface.co/chenghao)
### Licensing Information
Provide the license and link to the license webpage if available.
[CC BY-NC-SA 3.0](https://creativecommons.org/licenses/by-nc-sa/3.0/)
### Contributions
NA |
cheulyop | null | @article{bang2020ksponspeech,
title={KsponSpeech: Korean spontaneous speech corpus for automatic speech recognition},
author={Bang, Jeong-Uk and Yun, Seung and Kim, Seung-Hi and Choi, Mu-Yeol and Lee, Min-Kyu and Kim, Yeo-Jeong and Kim, Dong-Hyun and Park, Jun and Lee, Young-Jik and Kim, Sang-Hun},
journal={Applied Sciences},
volume={10},
number={19},
pages={6936},
year={2020},
publisher={Multidisciplinary Digital Publishing Institute}
} | KsponSpeech is a large-scale spontaneous speech corpus of Korean conversations. This corpus contains 969 hrs of general open-domain dialog utterances, spoken by about 2,000 native Korean speakers in a clean environment. All data were constructed by recording the dialogue of two people freely conversing on a variety of topics and manually transcribing the utterances. The transcription provides a dual transcription consisting of orthography and pronunciation, and disfluency tags for spontaneity of speech, such as filler words, repeated words, and word fragments. KsponSpeech is publicly available on an open data hub site of the Korea government. (https://aihub.or.kr/aidata/105) | false | 327 | false | cheulyop/ksponspeech | 2021-10-02T04:27:13.000Z | null | false | d51bd8aa4dcb0d95600de289e7c6ea761d412c2d | [] | [] | https://huggingface.co/datasets/cheulyop/ksponspeech/resolve/main/README.md | ---
YAML tags:
- copy-paste the tags obtained with the tagging app: https://github.com/huggingface/datasets-tagging
---
# Dataset Card for [KsponSpeech]
## 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
KsponSpeech is a large-scale spontaneous speech corpus of Korean conversations. This corpus contains 969 hrs of general open-domain dialog utterances, spoken by about 2,000 native Korean speakers in a clean environment. All data were constructed by recording the dialogue of two people freely conversing on a variety of topics and manually transcribing the utterances. The transcription provides a dual transcription consisting of orthography and pronunciation, and disfluency tags for spontaneity of speech, such as filler words, repeated words, and word fragments. KsponSpeech is publicly available on an open data hub site of the Korea government. (https://aihub.or.kr/aidata/105)
### 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.
|
chitra | null | null | null | false | 322 | false | chitra/contradictionNLI | 2021-12-29T10:45:19.000Z | null | false | b1e632ba5e39486891c9ade0d6ba70561993c91d | [] | [] | https://huggingface.co/datasets/chitra/contradictionNLI/resolve/main/README.md | This data can help in solving contradiction detection problem. this data is picked from kaggle.
reference - Contradictory, My DWatson |
chmanoj | null | null | null | false | 320 | false | chmanoj/ai4bharat__samanantar_processed_te | 2022-02-05T04:02:51.000Z | null | false | de949e03d6bdecb42f9300fb9be8f5a9b5acf5f4 | [] | [] | https://huggingface.co/datasets/chmanoj/ai4bharat__samanantar_processed_te/resolve/main/README.md | This is extracted from telugu subset from https://huggingface.co/datasets/ai4bharat/samanantar - used to create telugu kenLM models for ASR decoding. |
chopey | null | null | null | false | 321 | false | chopey/dhivehi | 2021-11-30T03:41:11.000Z | null | false | 24fba98c601fcde47d5a50fe72d54fdf70b69e11 | [] | [] | https://huggingface.co/datasets/chopey/dhivehi/resolve/main/README.md | Dhivehi dataset for MNT |
clarin-pl | null | """
_DESCRIPTION = | This dataset is designed to be used in training models
that restore punctuation marks from the output of
Automatic Speech Recognition system for Polish language. | false | 319 | false | clarin-pl/2021-punctuation-restoration | 2022-08-29T16:39:18.000Z | null | false | 6051cc3ed097fbbe93c7cc2c480279e230f43e93 | [] | [
"annotations_creators:crowdsourced",
"language:pl",
"language_creators:crowdsourced",
"multilinguality:monolingual",
"size_categories:n<1K",
"task_categories:automatic-speech-recognition"
] | https://huggingface.co/datasets/clarin-pl/2021-punctuation-restoration/resolve/main/README.md | ---
annotations_creators:
- crowdsourced
language:
- pl
language_creators:
- crowdsourced
license: []
multilinguality:
- monolingual
pretty_name: 2021-punctuation-restoration
size_categories:
- n<1K
source_datasets: []
tags: []
task_categories:
- automatic-speech-recognition
task_ids: []
---
# Punctuation restoration from read text
Restore punctuation marks from the output of an ASR system.
## Motivation
Speech transcripts generated by Automatic Speech Recognition (ASR) systems typically do not contain any punctuation or capitalization. In longer stretches of automatically recognized speech, the lack of punctuation affects the general clarity of the output text [1]. The primary purpose of punctuation (PR) and capitalization restoration (CR) as a distinct natural language processing (NLP) task is to improve the legibility of ASR-generated text, and possibly other types of texts without punctuation. Aside from their intrinsic value, PR and CR may improve the performance of other NLP aspects such as Named Entity Recognition (NER), part-of-speech (POS) and semantic parsing or spoken dialog segmentation [2, 3]. As useful as it seems, It is hard to systematically evaluate PR on transcripts of conversational language; mainly because punctuation rules can be ambiguous even for originally written texts, and the very nature of naturally-occurring spoken language makes it difficult to identify clear phrase and sentence boundaries [4,5]. Given these requirements and limitations, a PR task based on a redistributable corpus of read speech was suggested. 1200 texts included in this collection (totaling over 240,000 words) were selected from two distinct sources: WikiNews and WikiTalks. Punctuation found in these sources should be approached with some reservation when used for evaluation: these are original texts and may contain some user-induced errors and bias. The texts were read out by over a hundred different speakers. Original texts with punctuation were forced-aligned with recordings and used as the ideal ASR output. The goal of the task is to provide a solution for restoring punctuation in the test set collated for this task. The test set consists of time-aligned ASR transcriptions of read texts from the two sources. Participants are encouraged to use both text-based and speech-derived features to identify punctuation symbols (e.g. multimodal framework [6]). In addition, the train set is accompanied by reference text corpora of WikiNews and WikiTalks data that can be used in training and fine-tuning punctuation models.
## Task description
The purpose of this task is to restore punctuation in the ASR recognition of texts read out loud.
**Input** ('tokens*'* column): sequence of tokens
**Output** ('tags*'* column): sequence of tags
**Measurements**: F1-score (seqeval)
**Example**:
Input: `['selekcjoner', 'szosowej', 'kadry', 'elity', 'mężczyzn', 'piotr', 'wadecki', 'ogłosił', '27', 'marca', '2008', 'r', 'szeroki', 'skład', 'zawodników', 'którzy', 'będą', 'rywalizować', 'o', 'miejsce', 'w', 'reprezentacji', 'na', 'tour', 'de', 'pologne', 'lista', 'liczy', '22', 'nazwiska', 'zawodników', 'zarówno', 'z', 'zagranicznych', 'jaki', 'i', 'polskich', 'ekip', 'spośród', '22', 'wybrańców', 'selekcjonera', 'do', 'składu', 'dostanie', 'się', 'tylko', 'ośmiu', 'kolarzy', 'którzy', 'we', 'wrześniu', 'będą', 'rywalizować', 'z', 'najlepszymi', 'grupami', 'kolarskimi', 'na', 'świecie', 'w', 'kręgu', 'zainteresowania', 'wadeckiego', 'znajduje', 'się', 'także', 'pięciu', 'innych', 'zawodników', 'ale', 'oni', 'prawdopodobnie', 'wystartują', 'w', 'polskim', 'tourze', 'w', 'szeregach', 'swoich', 'ekip', 'szeroka', 'kadra', 'na', 'tour', 'de', 'pologne', 'dariusz', 'baranowski', 'łukasz', 'bodnar', 'bartosz', 'huzarski', 'błażej', 'janiaczyk', 'tomasz', 'kiendyś', 'mateusz', 'komar', 'tomasz', 'lisowicz', 'piotr', 'mazur', 'jacek', 'morajko', 'przemysław', 'niemiec', 'marek', 'rutkiewicz', 'krzysztof', 'szczawiński', 'mateusz', 'taciak', 'adam', 'wadecki', 'mariusz', 'witecki', 'piotr', 'zaradny', 'piotr', 'zieliński', 'mateusz', 'mróz', 'marek', 'wesoły', 'jarosław', 'rębiewski', 'robert', 'radosz', 'jarosław', 'dąbrowski']`
Input (translated by DeepL): `the selector of the men's elite road cycling team piotr wadecki announced on march 27, 2008 a wide line-up of riders who will compete for a place in the national team for the tour de pologne the list includes 22 names of riders both from foreign and Polish teams out of the 22 selected by the selector only eight riders will get into the line-up who in September will compete with the best cycling groups in the world wadecki's circle of interest also includes five other cyclists, but they will probably compete in the Polish tour in the ranks of their teams wide cadre for the tour de pologne dariusz baranowski łukasz bodnar bartosz huzarski błażej janiaczyk tomasz kiendyś mateusz komar tomasz lisowicz piotr mazur jacek morajko przemysław german marek rutkiewicz krzysztof szczawiński mateusz taciak adam wadecki mariusz witecki piotr zaradny piotr zieliński mateusz mróz marek wesoły jarosław rębiewski robert radosz jarosław dąbrowski`
Output: `['O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'B-.', 'O', 'O', 'B-,', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'B-.', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'B-,', 'O', 'O', 'O', 'B-.', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'B-,', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'B-.', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'B-,', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'B-.', 'O', 'O', 'O', 'O', 'O', 'B-:', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O']`
## Dataset – WikiPunct
WikiPunct is a crowdsourced text and audio data set of Polish Wikipedia pages read out loud by Polish lectors. The dataset is divided into two parts:conversational(WikiTalks)and information (WikiNews). Over a hundred people were involved in the production of the audio component. The total length of audio data reaches almost thirty-six hours, including the test set. Steps were taken to balance the male-to-female ratio.
WikiPuncthas over thirty-two thousand texts and 1200 audio files, one thousand in the training set and two hundred in the test set. There is a transcript of automatically recognized speech and force-aligned text for each text. The details behind the data format and evaluation metrics are presented below in the respective sections.
**Statistics:**
- **Text:**
- ver thirty-two thousand texts; WikiNews ca. 15,000, WikiTalks ca. 17,000;
- **Audio:**
- Selection procedure:
- randomly selected WikiNews (80% that is equal 800 entries for the training set) with the word count above 150 words and smaller than 300 words;
- randomly selected WikiTalks (20%) with word the count above 150 words but smaller than 300 words and at least one question mark
- Data set split
- Training data: 1000 recordings
- Test data: at 274 recordings
- Speakers:
- Polish male: 51 speakers, 16.7 hours of speech
- Polish female: 54 speakers, 19 hours of speech
**Data splits**
| Subset | Cardinality (texts) |
| ----------- | ----------------------: |
| train | 800 |
| dev | 0 |
| test | 200 |
**Class distribution (without "O")**
| Class | train | validation | test |
|:--------|--------:|-------------:|-------:|
| B-. | 0.419 | - | 0.416 |
| B-, | 0.406 | - | 0.403 |
| B-- | 0.097 | - | 0.099 |
| B-: | 0.037 | - | 0.052 |
| B-? | 0.032 | - | 0.024 |
| B-! | 0.005 | - | 0.004 |
| B-; | 0.004 | - | 0.002 |
**Punctuation for raw text:**
| | **symbol** | **mean** | **median** | **max** | **sum** | **included** |
| --- | --- | --- | --- | --- | --- | --- |
| **fullstop** | . | 12.44 | 7.0 | 1129.0 | 404 378 | yes |
| **comma** | , | 10.97 | 5.0 | 1283.0 | 356 678 | yes |
| **question\_mark** | ? | 0.83 | 0.0 | 130.0 | 26 879 | yes |
| **exclamation\_mark** | ! | 0.22 | 0.0 | 55.0 | 7 164 | yes |
| **hyphen** | - | 2.64 | 1.0 | 363.0 | 81 190 | yes |
| **colon** | : | 1.49 | 0.0 | 202.0 | 44 995 | yes |
| **ellipsis** | ... | 0.27 | 0.0 | 60.0 | 8 882 | yes |
| **semicolon** | ; | 0.13 | 0.0 | 51.0 | 4 270 | no |
| **quote** | " | 3.64 | 0.0 | 346.0 | 116 874 | no |
| **words** | | 169.50 | 89.0 | 17252.0 | 5 452 032 | - |
The dataset is divided into two parts: conversational (WikiTalks) and information (WikiNews).
**Part 1. WikiTalks**
Data scraped from Polish Wikipedia Talk pages. Talk pages, also known as discussion pages, are administration pages with editorial details and discussions for Wikipedia articles.. Talk pages were scrapped from the web using a list of article titles shared alongside Wikipedia dump archives.
Wikipedia Talk pages serve as conversational data. Here, users communicate with each other by writing comments. Vocabulary and punctuation errors are expected. This data set covers 20% of the spoken data.
Example:
- **wikitalks001948:** Cóż za bzdury tu powypisywane! Fra Diavolo starał się nie dopuścić do upadku Republiki Partenopejskiej? Kto to wymyślił?! Człowiek ten był jednym z najżarliwszych wrogów francuskiej okupacji, a za zasługi w wypędzeniu Francuzów został mianowany pułkownikiem w królewskiej armii z prawdziwie królewską pensją. Bez niego wyzwolenie, nazywać to tak czy też nie, północnej części królestwa byłoby dużo trudniejsze, bo dysponował siłą kilku tysięcy sprawnych w boju i umiejętnie wziętych w karby rzezimieszków. Toteż armia Burbonów nie pokonywała go, jak to się twierdzi w artykule, lecz ściśle współpracowała. Redaktorów zachęcam do jak najszybszej korekty artykułu, bo aktualnie jest obrazą dla ambicji Wikipedii. 91.199.250.17
- **wikitalks008902:** Stare wątki w dyskusji przeniosłem do archiwum. Od prawie roku dyskusja w nich nie była kontynuowana. Sławek Borewicz
**Part 2. WikiNews**
**Wikinews** is a free-content news wiki and a project of the Wikimedia Foundation. The site works through collaborative journalism. The data was scraped directly from wikinews dump archive. The overall text quality is high, but vocabulary and punctuation errors may occur. This data set covers 80% of the spoken data.
Example:
- **wikinews222361:** Misja STS-127 promu kosmicznego Endeavour do Międzynarodowej Stacji Kosmicznej została przełożona ze względu na wyciek wodoru. Podczas procesu napełniania zewnętrznego zbiornika paliwem, część ciekłego wodoru przemieniła się w gaz i przedostała się do systemu odpowietrzania. System ten jest używany do bezpiecznego odprowadzania nadmiaru wodoru z platformy startowej 39A do Centrum Lotów Kosmicznych imienia Johna F. Kennedy'ego. Początek misji miał mieć miejsce dzisiaj, o godzinie 13:17. Ze względu jednak na awarię, najbliższa możliwa data startu wahadłowca to środa 17 czerwca, jednak na ten dzień NASA na Przylądku Canaveral zaplanowana wystrzelenie sondy kosmicznej Lunar Reconnaissance Orbiter. Misja może być zatem opóźniona do 20 czerwca, który jest ostatnią możliwą datą startu w tym miesiącu. W niedzielę odbędzie się spotkanie specjalistów NASA, na którym zostanie ustalona nowa data startu i dalszy plan misji STS-127.
## Data format
Input is a TSV file with two columns:
1. Text ID (to be used when handling forced-aligned transcriptions and WAV files if needed)
2. Input text - in lower-case letter without punctuation marks
The output should have the same number of lines as the input file, in each line
the text with punctuation marks should be given.
### Forced-aligned transcriptions
We use force-aligned transcriptions of the original texts to approximate ASR output. Files in the _.clntmstmp_ format contain forced-alignment of the original text together with the audio file read out by a group of volunteers. The files may contain errors resulting from incorrect reading of the text (skipping fragments, adding words missing from the original text) and alignment errors resulting from the configuration of the alignment tool for text and audio files. The configuration targeted Polish; names from foreign languages may be poorly recognised, with the word duration equal to zero (start and end timestamps are equal). Data is given in the following format:
**(timestamp\_start,timestamp\_end) word**
...
**\</s\>**
where **\</s\>** is a symbol of the end of recognition.
Example:
(990,1200) Rosja
(1230,1500) zaczyna
(1590,1950) powracać
(1980,2040) do
(2070,2400) praktyk
(2430,2490) z
(2520,2760) czasów
(2820,3090) zimnej
(3180,3180) wojny.
(3960,4290) Rosjanie
(4380,4770) wznowili
(4860,5070) bowiem
(5100,5160) na
(5220,5430) stałe
(5520,5670) loty
(5760,6030) swoich
(6120,6600) bombowców
(6630,7230) strategicznych
(7350,7530) poza
(7590,7890) granice
(8010,8010) kraju.
(8880,9300) Prezydent
(9360,9810) Władimir
(9930,10200) Putin
(10650,10650) wyjaśnił,
(10830,10920) iż
(10980,11130) jest
(11160,11190) to
(11220,11520) odpowiedź
(11550,11640) na
(11670,12120) zagrożenie
(12240,12300) ze
(12330,12570) strony
(12660,12870) innych
(13140,13140) państw.
\</s\>
## Evaluation procedure
Baseline results will be provided in final evaluation.
### Punctuation
During the task the following punctuation marks will be evaluated:
| **Punctuation mark** | **symbol** |
| --- | --- |
| fullstop | . |
| comma | , |
| question mark | ? |
| exclamation mark | ! |
| hyphen | - |
| colon | : |
| ellipsis | ... |
| blank (no punctuation) | |
Note that semi-colon (`;`) is disregarded here.
### Submission format
The output to be evaluated is just the text with punctuation marks added.
### Metrics
Final results are evaluated in terms of precision, recall, and F1 scores for predicting each punctuation mark separately. Submissions are compared with respect to the weighted average of F1 scores for each punctuation mark.
##### Per-document score:
##### Global score per punctuation mark _p_:
Final scoring metric calculated as weighted average of global scores per
We would like to invite participants to discussion about evaluation metrics, taking into account such factors as:
- ASR and Forced-Alignment errors,
- inconsistencies among annotators,
- impact of only slight displacement of punctuation,
- assigning different weights to different types of errors.
### Video introduction
[](http://www.youtube.com/watch?v=yEh-RiFGN94 "Video instruction")
### Downloads
Data has been published in the following repository: https://github.com/poleval/2021-punctuation-restoration
Training data is provided in train/\*.tsv. Additional data can be downloaded from Google Drive. Below is a list of file names along with a description of what they contain.
- [poleval\_fa.train.tar.gz](https://drive.google.com/file/d/1oBFjZPb5Hk4r_VW4G0HrVnGy7A7zmTpa/view?usp=sharing) - archive contains forced-alignment of the original text together with the audio file
- [poleval\_wav.train.tar.gz](https://drive.google.com/file/d/1b6MyyqgA9D1U7DX3Vtgda7f9ppkxjCXJ/view?usp=sharing) - archive contains training audio files
- [poleval\_wav.validation.tar.gz](https://drive.google.com/file/d/1gwQRvrUtFqz3xGnmEN8znAzkBwC12Czu/view?usp=sharing) - archive contains test audio files
- [poleval\_text.rest.tar.gz](https://drive.google.com/file/d/10SdpLHPLXVfhJsq1okgC5fcxbFzCGoR5/view?usp=sharing) - archive contains additional text provided in JSON formatand CSV for which no audio files were provided (can be used for training purposes)
### Challenge stage
The competition in September 2021. Now the challenge is in the after-competition stage. You can submit solutions,
but they will be marked with a different color.
### License
Creative Commons - Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)
### References
1. Yi, J., Tao, J., Bai, Y., Tian, Z., & Fan, C. (2020). Adversarial transfer learning for punctuation restoration. _arXiv preprint arXiv:2004.00248_.
2. Nguyen, Thai Binh, et al. "Improving Vietnamese Named Entity Recognition from Speech Using Word Capitalization and Punctuation Recovery Models." _Proc. Interspeech 2020_ (2020): 4263-4267.
3. Hlubík, Pavel, et al. "Inserting Punctuation to ASR Output in a Real-Time Production Environment." _International Conference on Text, Speech, and Dialogue_. Springer, Cham, 2020.
4. Sirts, Kairit, and Kairit Peekman. "Evaluating Sentence Segmentation and Word Tokenization Systems on Estonian Web Texts." _Human Language Technologies–The Baltic Perspective: Proceedings of the Ninth International Conference Baltic HLT 2020_. Vol. 328. IOS Press, 2020.
5. Wang, Xueyujie. "Analysis of Sentence Boundary of the Host's Spoken Language Based on Semantic Orientation Pointwise Mutual Information Algorithm." _2020 12th International Conference on Measuring Technology and Mechatronics Automation (ICMTMA)_. IEEE, 2020.
6. Sunkara, Monica, et al. "Multimodal Semi-supervised Learning Framework for Punctuation Prediction in Conversational Speech." _arXiv preprint arXiv:2008.00702_ (2020).
|
clarin-pl | null | @misc{11321/849,
title = {{AspectEmo} 1.0: Multi-Domain Corpus of Consumer Reviews for Aspect-Based Sentiment Analysis},
author = {Koco{\'n}, Jan and Radom, Jarema and Kaczmarz-Wawryk, Ewa and Wabnic, Kamil and Zaj{\c a}czkowska, Ada and Za{\'s}ko-Zieli{\'n}ska, Monika},
url = {http://hdl.handle.net/11321/849},
note = {{CLARIN}-{PL} digital repository},
copyright = {The {MIT} License},
year = {2021}
} | AspectEmo dataset: Multi-Domain Corpus of Consumer Reviews for Aspect-Based
Sentiment Analysis | false | 355 | false | clarin-pl/aspectemo | 2022-08-29T16:39:32.000Z | null | false | 55467c09094ac3a0d8261013f884f8f3247b53a0 | [] | [
"annotations_creators:expert-generated",
"language_creators:other",
"language:pl",
"license:mit",
"multilinguality:monolingual",
"size_categories:1K",
"size_categories:1K<n<10K",
"source_datasets:original",
"task_categories:token-classification",
"task_ids:sentiment-classification"
] | https://huggingface.co/datasets/clarin-pl/aspectemo/resolve/main/README.md | ---
annotations_creators:
- expert-generated
language_creators:
- other
language:
- pl
license:
- mit
multilinguality:
- monolingual
pretty_name: 'AspectEmo'
size_categories:
- 1K
- 1K<n<10K
source_datasets:
- original
task_categories:
- token-classification
task_ids:
- sentiment-classification
---
# AspectEmo
## Description
AspectEmo Corpus is an extended version of a publicly available PolEmo 2.0 corpus of Polish customer reviews used in many projects on the use of different methods in sentiment analysis. The AspectEmo corpus consists of four subcorpora, each containing online customer reviews from the following domains: school, medicine, hotels, and products. All documents are annotated at the aspect level with six sentiment categories: strong negative (minus_m), weak negative (minus_s), neutral (zero), weak positive (plus_s), strong positive (plus_m).
## Versions
| version | config name | description | default | notes |
|---------|-------------|--------------------------------|---------|------------------|
| 1.0 | "1.0" | The version used in the paper. | YES | |
| 2.0 | - | Some bugs fixed. | NO | work in progress |
## Tasks (input, output and metrics)
Aspect-based sentiment analysis (ABSA) is a text analysis method that categorizes data by aspects and identifies the sentiment assigned to each aspect. It is the sequence tagging task.
**Input** ('*tokens'* column): sequence of tokens
**Output** ('*labels'* column): sequence of predicted tokens’ classes ("O" + 6 possible classes: strong negative (a_minus_m), weak negative (a_minus_s), neutral (a_zero), weak positive (a_plus_s), strong positive (a_plus_m), ambiguous (a_amb) )
**Domain**: school, medicine, hotels and products
**Measurements**: F1-score (seqeval)
**Example***:*
Input: `['Dużo', 'wymaga', ',', 'ale', 'bardzo', 'uczciwy', 'i', 'przyjazny', 'studentom', '.', 'Warto', 'chodzić', 'na', 'konsultacje', '.', 'Docenia', 'postępy', 'i', 'zaangażowanie', '.', 'Polecam', '.']`
Input (translated by DeepL): `'Demands a lot , but very honest and student friendly . Worth going to consultations . Appreciates progress and commitment . I recommend .'`
Output: `['O', 'a_plus_s', 'O', 'O', 'O', 'a_plus_m', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'a_zero', 'O', 'a_plus_m', 'O', 'O', 'O', 'O', 'O', 'O']`
## Data splits
| Subset | Cardinality (sentences) |
|:-------|------------------------:|
| train | 1173 |
| val | 0 |
| test | 292 |
## Class distribution(without "O")
| Class | train | validation | test |
|:----------|--------:|-------------:|-------:|
| a_plus_m | 0.359 | - | 0.369 |
| a_minus_m | 0.305 | - | 0.377 |
| a_zero | 0.234 | - | 0.182 |
| a_minus_s | 0.037 | - | 0.024 |
| a_plus_s | 0.037 | - | 0.015 |
| a_amb | 0.027 | - | 0.033 |
## Citation
```
@misc{11321/849,
title = {{AspectEmo} 1.0: Multi-Domain Corpus of Consumer Reviews for Aspect-Based Sentiment Analysis},
author = {Koco{\'n}, Jan and Radom, Jarema and Kaczmarz-Wawryk, Ewa and Wabnic, Kamil and Zaj{\c a}czkowska, Ada and Za{\'s}ko-Zieli{\'n}ska, Monika},
url = {http://hdl.handle.net/11321/849},
note = {{CLARIN}-{PL} digital repository},
copyright = {The {MIT} License},
year = {2021}
}
```
## License
```
The MIT License
```
## Links
[HuggingFace](https://huggingface.co/datasets/clarin-pl/aspectemo)
[Source](https://clarin-pl.eu/dspace/handle/11321/849)
[Paper](https://sentic.net/sentire2021kocon.pdf)
## Examples
### Loading
```python
from pprint import pprint
from datasets import load_dataset
dataset = load_dataset("clarin-pl/aspectemo")
pprint(dataset['train'][20])
# {'labels': [0, 4, 0, 0, 0, 5, 0, 0, 0, 0, 0, 0, 0, 3, 0, 5, 0, 0, 0, 0, 0, 0],
# 'tokens': ['Dużo',
# 'wymaga',
# ',',
# 'ale',
# 'bardzo',
# 'uczciwy',
# 'i',
# 'przyjazny',
# 'studentom',
# '.',
# 'Warto',
# 'chodzić',
# 'na',
# 'konsultacje',
# '.',
# 'Docenia',
# 'postępy',
# 'i',
# 'zaangażowanie',
# '.',
# 'Polecam',
# '.']}
```
### Evaluation
```python
import random
from pprint import pprint
from datasets import load_dataset, load_metric
dataset = load_dataset("clarin-pl/aspectemo")
references = dataset["test"]["labels"]
# generate random predictions
predictions = [
[
random.randrange(dataset["train"].features["labels"].feature.num_classes)
for _ in range(len(labels))
]
for labels in references
]
# transform to original names of labels
references_named = [
[dataset["train"].features["labels"].feature.names[label] for label in labels]
for labels in references
]
predictions_named = [
[dataset["train"].features["labels"].feature.names[label] for label in labels]
for labels in predictions
]
# transform to BILOU scheme
references_named = [
[f"U-{label}" if label != "O" else label for label in labels]
for labels in references_named
]
predictions_named = [
[f"U-{label}" if label != "O" else label for label in labels]
for labels in predictions_named
]
# utilise seqeval to evaluate
seqeval = load_metric("seqeval")
seqeval_score = seqeval.compute(
predictions=predictions_named,
references=references_named,
scheme="BILOU",
mode="strict",
)
pprint(seqeval_score)
# {'a_amb': {'f1': 0.00597237775289287,
# 'number': 91,
# 'precision': 0.003037782418834251,
# 'recall': 0.17582417582417584},
# 'a_minus_m': {'f1': 0.048306148055207034,
# 'number': 1039,
# 'precision': 0.0288551620760727,
# 'recall': 0.1482194417709336},
# 'a_minus_s': {'f1': 0.004682997118155619,
# 'number': 67,
# 'precision': 0.0023701002734731083,
# 'recall': 0.19402985074626866},
# 'a_plus_m': {'f1': 0.045933014354066985,
# 'number': 1015,
# 'precision': 0.027402473834443386,
# 'recall': 0.14187192118226602},
# 'a_plus_s': {'f1': 0.0021750951604132683,
# 'number': 41,
# 'precision': 0.001095690284879474,
# 'recall': 0.14634146341463414},
# 'a_zero': {'f1': 0.025159400310184387,
# 'number': 501,
# 'precision': 0.013768389287061486,
# 'recall': 0.14570858283433133},
# 'overall_accuracy': 0.13970115681233933,
# 'overall_f1': 0.02328248652368391,
# 'overall_precision': 0.012639312620633834,
# 'overall_recall': 0.14742193173565724}
``` |
clarin-pl | null | null | KPWR-NER tagging dataset. | false | 1,248 | false | clarin-pl/kpwr-ner | 2022-08-29T16:39:44.000Z | null | false | 6fd17a22c100eb9039060e986cc5b97d2831fdab | [] | [
"annotations_creators:expert-generated",
"language_creators:found",
"language:pl",
"license:cc-by-3.0",
"multilinguality:monolingual",
"size_categories:18K",
"size_categories:10K<n<100K",
"source_datasets:original",
"task_ids:named-entity-recognition"
] | https://huggingface.co/datasets/clarin-pl/kpwr-ner/resolve/main/README.md | ---
annotations_creators:
- expert-generated
language_creators:
- found
language:
- pl
license:
- cc-by-3.0
multilinguality:
- monolingual
pretty_name: 'KPWr-NER'
size_categories:
- 18K
- 10K<n<100K
source_datasets:
- original
task_categories:
- structure-prediction
task_ids:
- named-entity-recognition
---
# KPWR-NER
## Description
KPWR-NER is a part the Polish Corpus of Wrocław University of Technology (*Korpus Języka Polskiego Politechniki Wrocławskiej*). Its objective is named entity recognition for fine-grained categories of entities. It is the ‘n82’ version of the KPWr, which means that number of classes is restricted to 82 (originally 120). During corpus creation, texts were annotated by humans from various sources, covering many domains and genres.
## Tasks (input, output and metrics)
Named entity recognition (NER) - tagging entities in text with their corresponding type.
**Input** ('*tokens'* column): sequence of tokens
**Output** ('*ner'* column): sequence of predicted tokens’ classes in BIO notation (82 possible classes, described in detail in the annotation guidelines)
**Measurements**: F1-score (seqeval)
**Example**:
Input: `[‘Roboty’, ‘mają’, ‘kilkanaście’, ‘lat’, ‘i’, ‘pochodzą’, ‘z’, ‘USA’, ‘,’, ‘Wysokie’, ‘napięcie’, ‘jest’, ‘dużo’, ‘młodsze’, ‘,’, ‘powstało’, ‘w’, ‘Niemczech’, ‘.’]`
Input (translated by DeepL): `Robots are more than a dozen years old and come from the US, High Voltage is much younger, having been developed in Germany.`
Output: `[‘B-nam_pro_title’, ‘O’, ‘O’, ‘O’, ‘O’, ‘O’, ‘O’, ‘B-nam_loc_gpe_country’, ‘O’, ‘B-nam_pro_title’, ‘I-nam_pro_title’, ‘O’, ‘O’, ‘O’, ‘O’, ‘O’, ‘O’, ‘B-nam_loc_gpe_country’, ‘O’]`
## Data splits
| Subset | Cardinality (sentences) |
|--------|------------------------:|
| train | 13959 |
| dev | 0 |
| test | 4323 |
## Class distribution (without "O" and "I-*")
| Class | train | validation | test |
|:----------------------------|--------:|-------------:|----------:|
| B-nam_liv_person | 0.21910 | - | 0.21422 |
| B-nam_loc_gpe_city | 0.10101 | - | 0.09865 |
| B-nam_loc_gpe_country | 0.07467 | - | 0.08059 |
| B-nam_org_institution | 0.05893 | - | 0.06005 |
| B-nam_org_organization | 0.04448 | - | 0.05553 |
| B-nam_org_group_team | 0.03492 | - | 0.03363 |
| B-nam_adj_country | 0.03410 | - | 0.03747 |
| B-nam_org_company | 0.02439 | - | 0.01716 |
| B-nam_pro_media_periodic | 0.02250 | - | 0.01896 |
| B-nam_fac_road | 0.01995 | - | 0.02144 |
| B-nam_liv_god | 0.01934 | - | 0.00790 |
| B-nam_org_nation | 0.01739 | - | 0.01828 |
| B-nam_oth_tech | 0.01724 | - | 0.01377 |
| B-nam_pro_media_web | 0.01709 | - | 0.00903 |
| B-nam_fac_goe | 0.01596 | - | 0.01445 |
| B-nam_eve_human | 0.01573 | - | 0.01761 |
| B-nam_pro_title | 0.01558 | - | 0.00790 |
| B-nam_pro_brand | 0.01543 | - | 0.01038 |
| B-nam_org_political_party | 0.01264 | - | 0.01309 |
| B-nam_loc_gpe_admin1 | 0.01219 | - | 0.01445 |
| B-nam_eve_human_sport | 0.01174 | - | 0.01242 |
| B-nam_pro_software | 0.01091 | - | 0.02190 |
| B-nam_adj | 0.00963 | - | 0.01174 |
| B-nam_loc_gpe_admin3 | 0.00888 | - | 0.01061 |
| B-nam_pro_model_car | 0.00873 | - | 0.00587 |
| B-nam_loc_hydronym_river | 0.00843 | - | 0.01151 |
| B-nam_oth | 0.00775 | - | 0.00497 |
| B-nam_pro_title_document | 0.00738 | - | 0.01986 |
| B-nam_loc_astronomical | 0.00730 | - | - |
| B-nam_oth_currency | 0.00723 | - | 0.01151 |
| B-nam_adj_city | 0.00670 | - | 0.00948 |
| B-nam_org_group_band | 0.00587 | - | 0.00429 |
| B-nam_loc_gpe_admin2 | 0.00565 | - | 0.00813 |
| B-nam_loc_gpe_district | 0.00504 | - | 0.00406 |
| B-nam_loc_land_continent | 0.00459 | - | 0.00722 |
| B-nam_loc_country_region | 0.00459 | - | 0.00090 |
| B-nam_loc_land_mountain | 0.00414 | - | 0.00203 |
| B-nam_pro_title_book | 0.00384 | - | 0.00248 |
| B-nam_loc_historical_region | 0.00376 | - | 0.00497 |
| B-nam_loc | 0.00361 | - | 0.00090 |
| B-nam_eve | 0.00361 | - | 0.00181 |
| B-nam_org_group | 0.00331 | - | 0.00406 |
| B-nam_loc_land_island | 0.00331 | - | 0.00248 |
| B-nam_pro_media_tv | 0.00316 | - | 0.00158 |
| B-nam_liv_habitant | 0.00316 | - | 0.00158 |
| B-nam_eve_human_cultural | 0.00316 | - | 0.00497 |
| B-nam_pro_title_tv | 0.00309 | - | 0.00542 |
| B-nam_oth_license | 0.00286 | - | 0.00248 |
| B-nam_num_house | 0.00256 | - | 0.00248 |
| B-nam_pro_title_treaty | 0.00248 | - | 0.00045 |
| B-nam_fac_system | 0.00248 | - | 0.00587 |
| B-nam_loc_gpe_subdivision | 0.00241 | - | 0.00587 |
| B-nam_loc_land_region | 0.00226 | - | 0.00248 |
| B-nam_pro_title_album | 0.00218 | - | 0.00158 |
| B-nam_adj_person | 0.00203 | - | 0.00406 |
| B-nam_fac_square | 0.00196 | - | 0.00135 |
| B-nam_pro_award | 0.00188 | - | 0.00519 |
| B-nam_eve_human_holiday | 0.00188 | - | 0.00203 |
| B-nam_pro_title_song | 0.00166 | - | 0.00158 |
| B-nam_pro_media_radio | 0.00151 | - | 0.00068 |
| B-nam_pro_vehicle | 0.00151 | - | 0.00090 |
| B-nam_oth_position | 0.00143 | - | 0.00226 |
| B-nam_liv_animal | 0.00143 | - | 0.00248 |
| B-nam_pro | 0.00135 | - | 0.00045 |
| B-nam_oth_www | 0.00120 | - | 0.00451 |
| B-nam_num_phone | 0.00120 | - | 0.00045 |
| B-nam_pro_title_article | 0.00113 | - | - |
| B-nam_oth_data_format | 0.00113 | - | 0.00226 |
| B-nam_fac_bridge | 0.00105 | - | 0.00090 |
| B-nam_liv_character | 0.00098 | - | - |
| B-nam_pro_software_game | 0.00090 | - | 0.00068 |
| B-nam_loc_hydronym_lake | 0.00090 | - | 0.00045 |
| B-nam_loc_gpe_conurbation | 0.00090 | - | - |
| B-nam_pro_media | 0.00083 | - | 0.00181 |
| B-nam_loc_land | 0.00075 | - | 0.00045 |
| B-nam_loc_land_peak | 0.00075 | - | - |
| B-nam_fac_park | 0.00068 | - | 0.00226 |
| B-nam_org_organization_sub | 0.00060 | - | 0.00068 |
| B-nam_loc_hydronym | 0.00060 | - | 0.00023 |
| B-nam_loc_hydronym_sea | 0.00045 | - | 0.00068 |
| B-nam_loc_hydronym_ocean | 0.00045 | - | 0.00023 |
| B-nam_fac_goe_stop | 0.00038 | - | 0.00090 |
## Citation
```
@inproceedings{broda-etal-2012-kpwr,
title = "{KPW}r: Towards a Free Corpus of {P}olish",
author = "Broda, Bartosz and
Marci{\'n}czuk, Micha{\l} and
Maziarz, Marek and
Radziszewski, Adam and
Wardy{\'n}ski, Adam",
booktitle = "Proceedings of the Eighth International Conference on Language Resources and Evaluation ({LREC}'12)",
month = may,
year = "2012",
address = "Istanbul, Turkey",
publisher = "European Language Resources Association (ELRA)",
url = "http://www.lrec-conf.org/proceedings/lrec2012/pdf/965_Paper.pdf",
pages = "3218--3222",
abstract = "This paper presents our efforts aimed at collecting and annotating a free Polish corpus. The corpus will serve for us as training and testing material for experiments with Machine Learning algorithms. As others may also benefit from the resource, we are going to release it under a Creative Commons licence, which is hoped to remove unnecessary usage restrictions, but also to facilitate reproduction of our experimental results. The corpus is being annotated with various types of linguistic entities: chunks and named entities, selected syntactic and semantic relations, word senses and anaphora. We report on the current state of the project as well as our ultimate goals.",
}
```
## License
```
Creative Commons Attribution 3.0 Unported Licence
```
## Links
[HuggingFace](https://huggingface.co/datasets/clarin-pl/kpwr-ner)
[Source](https://clarin-pl.eu/index.php/kpwr-en/)
[Paper](https://aclanthology.org/L12-1574/)
[KPWr annotation guidelines](http://www.nlp.pwr.wroc.pl/narzedzia-i-zasoby/zasoby/kpwr-lemma/16-narzedzia-zasoby/79-wytyczne)
[KPWr annotation guidelines - named entities](https://clarin-pl.eu/dspace/handle/11321/294)
## Examples
### Loading
```python
from pprint import pprint
from datasets import load_dataset
dataset = load_dataset("clarin-pl/kpwr-ner")
pprint(dataset['train'][0])
# {'lemmas': ['roborally', 'czy', 'wysoki', 'napięcie', '?'],
# 'ner': [73, 160, 73, 151, 160],
# 'orth': ['subst:sg:nom:n',
# 'qub',
# 'adj:sg:nom:n:pos',
# 'subst:sg:nom:n',
# 'interp'],
# 'tokens': ['RoboRally', 'czy', 'Wysokie', 'napięcie', '?']}
```
### Evaluation
```python
import random
from pprint import pprint
from datasets import load_dataset, load_metric
dataset = load_dataset("clarin-pl/kpwr-ner")
references = dataset["test"]["ner"]
# generate random predictions
predictions = [
[
random.randrange(dataset["train"].features["ner"].feature.num_classes)
for _ in range(len(labels))
]
for labels in references
]
# transform to original names of labels
references_named = [
[dataset["train"].features["ner"].feature.names[label] for label in labels]
for labels in references
]
predictions_named = [
[dataset["train"].features["ner"].feature.names[label] for label in labels]
for labels in predictions
]
# utilise seqeval to evaluate
seqeval = load_metric("seqeval")
seqeval_score = seqeval.compute(
predictions=predictions_named, references=references_named, scheme="IOB2"
)
pprint(seqeval_score, depth=1)
# {'nam_adj': {...},
# 'nam_adj_city': {...},
# 'nam_adj_country': {...},
# 'nam_adj_person': {...},
# 'nam_eve': {...},
# 'nam_eve_human': {...},
# 'nam_eve_human_cultural': {...},
# 'nam_eve_human_holiday': {...},
# 'nam_eve_human_sport': {...},
# 'nam_fac_bridge': {...},
# 'nam_fac_goe': {...},
# 'nam_fac_goe_stop': {...},
# 'nam_fac_park': {...},
# 'nam_fac_road': {...},
# 'nam_fac_square': {...},
# 'nam_fac_system': {...},
# 'nam_liv_animal': {...},
# 'nam_liv_character': {...},
# 'nam_liv_god': {...},
# 'nam_liv_habitant': {...},
# 'nam_liv_person': {...},
# 'nam_loc': {...},
# 'nam_loc_astronomical': {...},
# 'nam_loc_country_region': {...},
# 'nam_loc_gpe_admin1': {...},
# 'nam_loc_gpe_admin2': {...},
# 'nam_loc_gpe_admin3': {...},
# 'nam_loc_gpe_city': {...},
# 'nam_loc_gpe_conurbation': {...},
# 'nam_loc_gpe_country': {...},
# 'nam_loc_gpe_district': {...},
# 'nam_loc_gpe_subdivision': {...},
# 'nam_loc_historical_region': {...},
# 'nam_loc_hydronym': {...},
# 'nam_loc_hydronym_lake': {...},
# 'nam_loc_hydronym_ocean': {...},
# 'nam_loc_hydronym_river': {...},
# 'nam_loc_hydronym_sea': {...},
# 'nam_loc_land': {...},
# 'nam_loc_land_continent': {...},
# 'nam_loc_land_island': {...},
# 'nam_loc_land_mountain': {...},
# 'nam_loc_land_peak': {...},
# 'nam_loc_land_region': {...},
# 'nam_num_house': {...},
# 'nam_num_phone': {...},
# 'nam_org_company': {...},
# 'nam_org_group': {...},
# 'nam_org_group_band': {...},
# 'nam_org_group_team': {...},
# 'nam_org_institution': {...},
# 'nam_org_nation': {...},
# 'nam_org_organization': {...},
# 'nam_org_organization_sub': {...},
# 'nam_org_political_party': {...},
# 'nam_oth': {...},
# 'nam_oth_currency': {...},
# 'nam_oth_data_format': {...},
# 'nam_oth_license': {...},
# 'nam_oth_position': {...},
# 'nam_oth_tech': {...},
# 'nam_oth_www': {...},
# 'nam_pro': {...},
# 'nam_pro_award': {...},
# 'nam_pro_brand': {...},
# 'nam_pro_media': {...},
# 'nam_pro_media_periodic': {...},
# 'nam_pro_media_radio': {...},
# 'nam_pro_media_tv': {...},
# 'nam_pro_media_web': {...},
# 'nam_pro_model_car': {...},
# 'nam_pro_software': {...},
# 'nam_pro_software_game': {...},
# 'nam_pro_title': {...},
# 'nam_pro_title_album': {...},
# 'nam_pro_title_article': {...},
# 'nam_pro_title_book': {...},
# 'nam_pro_title_document': {...},
# 'nam_pro_title_song': {...},
# 'nam_pro_title_treaty': {...},
# 'nam_pro_title_tv': {...},
# 'nam_pro_vehicle': {...},
# 'overall_accuracy': 0.006156203762418094,
# 'overall_f1': 0.0009844258777797407,
# 'overall_precision': 0.0005213624939842789,
# 'overall_recall': 0.008803611738148984}
``` |
clarin-pl | null | null | NKJP-POS tagging dataset. | false | 431 | false | clarin-pl/nkjp-pos | 2022-08-29T16:39:54.000Z | null | false | 03b3c3c98a06e64e47878ddfd67ae69f03bf2419 | [] | [
"annotations_creators:expert-generated",
"language_creators:other",
"language:pl",
"license:gpl-3.0",
"multilinguality:monolingual",
"size_categories:unknown",
"source_datasets:original",
"task_ids:part-of-speech-tagging"
] | https://huggingface.co/datasets/clarin-pl/nkjp-pos/resolve/main/README.md | ---
annotations_creators:
- expert-generated
language_creators:
- other
language:
- pl
license:
- gpl-3.0
multilinguality:
- monolingual
pretty_name: 'nkjp-pos'
size_categories:
- unknown
source_datasets:
- original
task_categories:
- structure-prediction
task_ids:
- part-of-speech-tagging
---
# nkjp-pos
## Description
NKJP-POS is a part the National Corpus of Polish (*Narodowy Korpus Języka Polskiego*). Its objective is part-of-speech tagging, e.g. nouns, verbs, adjectives, adverbs, etc. During the creation of corpus, texts of were annotated by humans from various sources, covering many domains and genres.
## Tasks (input, output and metrics)
Part-of-speech tagging (POS tagging) - tagging words in text with their corresponding part of speech.
**Input** ('*tokens'* column): sequence of tokens
**Output** ('*pos_tags'* column): sequence of predicted tokens’ classes (35 possible classes, described in detail in the annotation guidelines)
**Measurements**: F1-score (seqeval)
**Example***:*
Input: `['Zarejestruj', 'się', 'jako', 'bezrobotny', '.']`
Input (translated by DeepL): `Register as unemployed.`
Output: `['impt', 'qub', 'conj', 'subst', 'interp']`
## Data splits
| Subset | Cardinality (sentences) |
| ----------- | ----------------------: |
| train | 78219 |
| dev | 0 |
| test | 7444 |
## Class distribution
| Class | train | dev | test |
|:--------|--------:|------:|--------:|
| subst | 0.27345 | - | 0.27656 |
| interp | 0.18101 | - | 0.17944 |
| adj | 0.10611 | - | 0.10919 |
| prep | 0.09567 | - | 0.09547 |
| qub | 0.05670 | - | 0.05491 |
| fin | 0.04939 | - | 0.04648 |
| praet | 0.04409 | - | 0.04348 |
| conj | 0.03711 | - | 0.03724 |
| adv | 0.03512 | - | 0.03333 |
| inf | 0.01591 | - | 0.01547 |
| comp | 0.01476 | - | 0.01439 |
| num | 0.01322 | - | 0.01436 |
| ppron3 | 0.01111 | - | 0.01018 |
| ppas | 0.01086 | - | 0.01085 |
| ger | 0.00961 | - | 0.01050 |
| brev | 0.00856 | - | 0.01181 |
| ppron12 | 0.00670 | - | 0.00665 |
| aglt | 0.00629 | - | 0.00602 |
| pred | 0.00539 | - | 0.00540 |
| pact | 0.00454 | - | 0.00452 |
| bedzie | 0.00229 | - | 0.00243 |
| pcon | 0.00218 | - | 0.00189 |
| impt | 0.00203 | - | 0.00226 |
| siebie | 0.00177 | - | 0.00158 |
| imps | 0.00174 | - | 0.00177 |
| interj | 0.00131 | - | 0.00102 |
| xxx | 0.00070 | - | 0.00048 |
| adjp | 0.00069 | - | 0.00065 |
| winien | 0.00068 | - | 0.00057 |
| adja | 0.00048 | - | 0.00058 |
| pant | 0.00012 | - | 0.00018 |
| burk | 0.00011 | - | 0.00006 |
| numcol | 0.00011 | - | 0.00013 |
| depr | 0.00010 | - | 0.00004 |
| adjc | 0.00007 | - | 0.00008 |
## Citation
```
@book{przepiorkowski_narodowy_2012,
title = {Narodowy korpus języka polskiego},
isbn = {978-83-01-16700-4},
language = {pl},
publisher = {Wydawnictwo Naukowe PWN},
editor = {Przepiórkowski, Adam and Bańko, Mirosław and Górski, Rafał L. and Lewandowska-Tomaszczyk, Barbara},
year = {2012}
}
```
## License
```
GNU GPL v.3
```
## Links
[HuggingFace](https://huggingface.co/datasets/clarin-pl/nkjp-pos)
[Source](http://clip.ipipan.waw.pl/NationalCorpusOfPolish)
[Paper](http://nkjp.pl/settings/papers/NKJP_ksiazka.pdf)
## Examples
### Loading
```python
from pprint import pprint
from datasets import load_dataset
dataset = load_dataset("clarin-pl/nkjp-pos")
pprint(dataset['train'][5000])
# {'id': '130-2-900005_morph_49.49-s',
# 'pos_tags': [16, 4, 3, 30, 12, 18, 3, 16, 14, 6, 14, 26, 1, 30, 12],
# 'tokens': ['Najwyraźniej',
# 'źle',
# 'ocenił',
# 'odległość',
# ',',
# 'bo',
# 'zderzył',
# 'się',
# 'z',
# 'jadącą',
# 'z',
# 'naprzeciwka',
# 'ciężarową',
# 'scanią',
# '.']}
```
### Evaluation
```python
import random
from pprint import pprint
from datasets import load_dataset, load_metric
dataset = load_dataset("clarin-pl/nkjp-pos")
references = dataset["test"]["pos_tags"]
# generate random predictions
predictions = [
[
random.randrange(dataset["train"].features["pos_tags"].feature.num_classes)
for _ in range(len(labels))
]
for labels in references
]
# transform to original names of labels
references_named = [
[dataset["train"].features["pos_tags"].feature.names[label] for label in labels]
for labels in references
]
predictions_named = [
[dataset["train"].features["pos_tags"].feature.names[label] for label in labels]
for labels in predictions
]
# transform to BILOU scheme
references_named = [
[f"U-{label}" if label != "O" else label for label in labels]
for labels in references_named
]
predictions_named = [
[f"U-{label}" if label != "O" else label for label in labels]
for labels in predictions_named
]
# utilise seqeval to evaluate
seqeval = load_metric("seqeval")
seqeval_score = seqeval.compute(
predictions=predictions_named,
references=references_named,
scheme="BILOU",
mode="strict",
)
pprint(seqeval_score, depth=1)
# {'adj': {...},
# 'adja': {...},
# 'adjc': {...},
# 'adjp': {...},
# 'adv': {...},
# 'aglt': {...},
# 'bedzie': {...},
# 'brev': {...},
# 'burk': {...},
# 'comp': {...},
# 'conj': {...},
# 'depr': {...},
# 'fin': {...},
# 'ger': {...},
# 'imps': {...},
# 'impt': {...},
# 'inf': {...},
# 'interj': {...},
# 'interp': {...},
# 'num': {...},
# 'numcol': {...},
# 'overall_accuracy': 0.027855061488566583,
# 'overall_f1': 0.027855061488566583,
# 'overall_precision': 0.027855061488566583,
# 'overall_recall': 0.027855061488566583,
# 'pact': {...},
# 'pant': {...},
# 'pcon': {...},
# 'ppas': {...},
# 'ppron12': {...},
# 'ppron3': {...},
# 'praet': {...},
# 'pred': {...},
# 'prep': {...},
# 'qub': {...},
# 'siebie': {...},
# 'subst': {...},
# 'winien': {...},
# 'xxx': {...}}
``` |
clarin-pl | null | @inproceedings{kocon-etal-2019-multi,
title = "Multi-Level Sentiment Analysis of {P}ol{E}mo 2.0: Extended Corpus of Multi-Domain Consumer Reviews",
author = "Koco{\'n}, Jan and
Mi{\l}kowski, Piotr and
Za{\'s}ko-Zieli{\'n}ska, Monika",
booktitle = "Proceedings of the 23rd Conference on Computational Natural Language Learning (CoNLL)",
month = nov,
year = "2019",
address = "Hong Kong, China",
publisher = "Association for Computational Linguistics",
url = "https://www.aclweb.org/anthology/K19-1092",
doi = "10.18653/v1/K19-1092",
pages = "980--991",} | PolEmo 2.0: Corpus of Multi-Domain Consumer Reviews, evaluation data for article presented at CoNLL. | false | 2,959 | false | clarin-pl/polemo2-official | 2022-08-29T16:40:01.000Z | null | false | 802e35d2b12bae84bb07911d841e8f046dc2fcef | [] | [
"annotations_creators:expert-generated",
"language_creators:other",
"language:pl",
"license:cc-by-sa-4.0",
"multilinguality:monolingual",
"size_categories:8K",
"size_categories:1K<n<10K",
"source_datasets:original",
"task_categories:text-classification",
"task_ids:sentiment-classification"
] | https://huggingface.co/datasets/clarin-pl/polemo2-official/resolve/main/README.md | ---
annotations_creators:
- expert-generated
language_creators:
- other
language:
- pl
license:
- cc-by-sa-4.0
multilinguality:
- monolingual
pretty_name: 'Polemo2'
size_categories:
- 8K
- 1K<n<10K
source_datasets:
- original
task_categories:
- text-classification
task_ids:
- sentiment-classification
---
# Polemo2
## Description
The PolEmo2.0 is a dataset of online consumer reviews from four domains: medicine, hotels, products, and university. It is human-annotated on a level of full reviews and individual sentences. Current version (PolEmo 2.0) contains 8,216 reviews having 57,466 sentences. Each text and sentence was manually annotated with sentiment in the 2+1 scheme, which gives a total of 197,046 annotations. About 85% of the reviews are from the medicine and hotel domains. Each review is annotated with four labels: positive, negative, neutral, or ambiguous.
## Tasks (input, output and metrics)
The task is to predict the correct label of the review.
**Input** ('*text*' column): sentence
**Output** ('*target*' column): label for sentence sentiment ('zero': neutral, 'minus': negative, 'plus': positive, 'amb': ambiguous)
**Domain**: Online reviews
**Measurements**: Accuracy, F1 Macro
**Example**:
Input: `Na samym wejściu hotel śmierdzi . W pokojach jest pleśń na ścianach , brudny dywan . W łazience śmierdzi chemią , hotel nie grzeje w pokojach panuje chłód . Wyposażenie pokoju jest stare , kran się rusza , drzwi na balkon nie domykają się . Jedzenie jest w małych ilościach i nie smaczne . Nie polecam nikomu tego hotelu .`
Input (translated by DeepL): `At the very entrance the hotel stinks . In the rooms there is mold on the walls , dirty carpet . The bathroom smells of chemicals , the hotel does not heat in the rooms are cold . The room furnishings are old , the faucet moves , the door to the balcony does not close . The food is in small quantities and not tasty . I would not recommend this hotel to anyone .`
Output: `1` (negative)
## Data splits
| Subset | Cardinality |
|--------|------------:|
| train | 6573 |
| val | 823 |
| test | 820 |
## Class distribution
| Class | train | dev | test |
|:--------|--------:|-------------:|-------:|
| minus | 0.3756 | 0.3694 | 0.4134 |
| plus | 0.2775 | 0.2868 | 0.2768 |
| amb | 0.1991 | 0.1883 | 0.1659 |
| zero | 0.1477 | 0.1555 | 0.1439 |
## Citation
```
@inproceedings{kocon-etal-2019-multi,
title = "Multi-Level Sentiment Analysis of {P}ol{E}mo 2.0: Extended Corpus of Multi-Domain Consumer Reviews",
author = "Koco{\'n}, Jan and
Mi{\l}kowski, Piotr and
Za{\'s}ko-Zieli{\'n}ska, Monika",
booktitle = "Proceedings of the 23rd Conference on Computational Natural Language Learning (CoNLL)",
month = nov,
year = "2019",
address = "Hong Kong, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/K19-1092",
doi = "10.18653/v1/K19-1092",
pages = "980--991",
abstract = "In this article we present an extended version of PolEmo {--} a corpus of consumer reviews from 4 domains: medicine, hotels, products and school. Current version (PolEmo 2.0) contains 8,216 reviews having 57,466 sentences. Each text and sentence was manually annotated with sentiment in 2+1 scheme, which gives a total of 197,046 annotations. We obtained a high value of Positive Specific Agreement, which is 0.91 for texts and 0.88 for sentences. PolEmo 2.0 is publicly available under a Creative Commons copyright license. We explored recent deep learning approaches for the recognition of sentiment, such as Bi-directional Long Short-Term Memory (BiLSTM) and Bidirectional Encoder Representations from Transformers (BERT).",
}
```
## License
```
Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0)
```
## Links
[HuggingFace](https://huggingface.co/datasets/clarin-pl/polemo2-official)
[Source](https://clarin-pl.eu/dspace/handle/11321/710)
[Paper](https://aclanthology.org/K19-1092/)
## Examples
### Loading
```python
from pprint import pprint
from datasets import load_dataset
dataset = load_dataset("clarin-pl/polemo2-official")
pprint(dataset['train'][0])
# {'target': 1,
# 'text': 'Na samym wejściu hotel śmierdzi . W pokojach jest pleśń na ścianach '
# ', brudny dywan . W łazience śmierdzi chemią , hotel nie grzeje w '
# 'pokojach panuje chłód . Wyposażenie pokoju jest stare , kran się '
# 'rusza , drzwi na balkon nie domykają się . Jedzenie jest w małych '
# 'ilościach i nie smaczne . Nie polecam nikomu tego hotelu .'}
```
### Evaluation
```python
import random
from pprint import pprint
from datasets import load_dataset, load_metric
dataset = load_dataset("clarin-pl/polemo2-official")
references = dataset["test"]["target"]
# generate random predictions
predictions = [random.randrange(max(references) + 1) for _ in range(len(references))]
acc = load_metric("accuracy")
f1 = load_metric("f1")
acc_score = acc.compute(predictions=predictions, references=references)
f1_score = f1.compute(predictions=predictions, references=references, average='macro')
pprint(acc_score)
pprint(f1_score)
# {'accuracy': 0.2475609756097561}
# {'f1': 0.23747048177471738}
```
|
classla | null | @misc{ljubešić2019frenk,
title={The FRENK Datasets of Socially Unacceptable Discourse in Slovene and English},
author={Nikola Ljubešić and Darja Fišer and Tomaž Erjavec},
year={2019},
eprint={1906.02045},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/1906.02045}
} | The FRENK Datasets of Socially Unacceptable Discourse in English. | false | 489 | false | classla/FRENK-hate-en | 2022-10-21T07:52:06.000Z | null | false | 52483dba0ff23291271ee9249839865e3c3e7e50 | [] | [
"arxiv:1906.02045",
"language:en",
"license:other",
"size_categories:1K<n<10K",
"task_categories:text-classification",
"tags:hate-speech-detection",
"tags:offensive-language"
] | https://huggingface.co/datasets/classla/FRENK-hate-en/resolve/main/README.md | ---
language:
- en
license:
- other
size_categories:
- 1K<n<10K
task_categories:
- text-classification
task_ids: []
tags:
- hate-speech-detection
- offensive-language
---
# Offensive language dataset of Croatian comments FRENK 1.0
English subset of the [FRENK dataset](http://hdl.handle.net/11356/1433). Also available on HuggingFace dataset hub: [Croatian subset](https://huggingface.co/datasets/5roop/FRENK-hate-hr), [Slovenian subset](https://huggingface.co/datasets/5roop/FRENK-hate-sl).
## Dataset Description
- **Homepage:** http://hdl.handle.net/11356/1433
- **Repository:** http://hdl.handle.net/11356/1433
- **Paper:** https://arxiv.org/abs/1906.02045
- **Project page** https://nl.ijs.si/frenk/
## Description of the original dataset
The original FRENK dataset consists of comments to Facebook posts (news articles) of mainstream media outlets from Croatia, Great Britain, and Slovenia, on the topics of migrants and LGBT. The dataset contains whole discussion threads. Each comment is annotated by the type of socially unacceptable discourse (e.g., inappropriate, offensive, violent speech) and its target (e.g., migrants/LGBT, commenters, media). The annotation schema is described in detail in [https://arxiv.org/pdf/1906.02045.pdf]. Usernames in the metadata are pseudo-anonymised and removed from the comments.
The data in each language (Croatian (hr), English (en), Slovenian (sl), and topic (migrants, LGBT) is divided into a training and a testing portion. The training and testing data consist of separate discussion threads, i.e., there is no cross-discussion-thread contamination between training and testing data. The sizes of the splits are the following: Croatian, migrants: 4356 training comments, 978 testing comments; Croatian LGBT: 4494 training comments, 1142 comments; English, migrants: 4540 training comments, 1285 testing comments; English, LGBT: 4819 training comments, 1017 testing comments; Slovenian, migrants: 5145 training comments, 1277 testing comments; Slovenian, LGBT: 2842 training comments, 900 testing comments.
For this dataset only the English data was used. Training segment has been split into beginning 90% (published here as training split) and end 10% (published here as dev split).
## Usage in `Transformers`
```python
import datasets
ds = datasets.load_dataset("classla/FRENK-hate-en","binary")
```
For binary classification the following encoding is used:
```python
_CLASS_MAP_BINARY = {
'Acceptable': 0,
'Offensive': 1,
}
```
The original labels are available if the dataset is loaded with the `multiclass` option:
```python
import datasets
ds = datasets.load_dataset("5roop/FRENK-hate-en","multiclass").
```
In this case the encoding used is:
```python
_CLASS_MAP_MULTICLASS = {
'Acceptable speech': 0,
'Inappropriate': 1,
'Background offensive': 2,
'Other offensive': 3,
'Background violence': 4,
'Other violence': 5,
}
```
The original labels are available if the dataset is loaded with the `multiclass` option:
```python
import datasets
ds = datasets.load_dataset("classla/FRENK-hate-en","multiclass").
```
In this case the encoding used is:
```python
_CLASS_MAP_MULTICLASS = {
'Acceptable speech': 0,
'Inappropriate': 1,
'Background offensive': 2,
'Other offensive': 3,
'Background violence': 4,
'Other violence': 5,
}
```
## Data structure
* `text`: text
* `target`: who is the target of the hate-speech text ("no target", "commenter", "target" (migrants or LGBT, depending on the topic), or "related to" (again, the topic))
* `topic`: whether the text relates to lgbt or migrants hate-speech domains
* `label`: label of the text instance, see above.
## Data instance
```
{'text': "Not everyone has the option of a rainbow reaction; I don't but wish I did.",
'target': 'No target',
'topic': 'lgbt',
'label': 0}
```
## Licensing information
CLARIN.SI Licence ACA ID-BY-NC-INF-NORED 1.0
## Citation information
When using this dataset please cite the following paper:
```
@misc{ljubešić2019frenk,
title={The FRENK Datasets of Socially Unacceptable Discourse in Slovene and English},
author={Nikola Ljubešić and Darja Fišer and Tomaž Erjavec},
year={2019},
eprint={1906.02045},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/1906.02045}
}
```
The original dataset can be cited as
```
@misc{11356/1433,
title = {Offensive language dataset of Croatian, English and Slovenian comments {FRENK} 1.0},
author = {Ljube{\v s}i{\'c}, Nikola and Fi{\v s}er, Darja and Erjavec, Toma{\v z}},
url = {http://hdl.handle.net/11356/1433},
note = {Slovenian language resource repository {CLARIN}.{SI}},
copyright = {{CLARIN}.{SI} Licence {ACA} {ID}-{BY}-{NC}-{INF}-{NORED} 1.0},
year = {2021} }
```
|
classla | null | @misc{ljubešić2019frenk,
title={The FRENK Datasets of Socially Unacceptable Discourse in Slovene and English},
author={Nikola Ljubešić and Darja Fišer and Tomaž Erjavec},
year={2019},
eprint={1906.02045},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/1906.02045}
} | The FRENK Datasets of Socially Unacceptable Discourse in Croatian. | false | 483 | false | classla/FRENK-hate-hr | 2022-10-21T07:46:28.000Z | null | false | e7fc9f3d8d6c5640a26679d8a50b1666b02cc41f | [] | [
"arxiv:1906.02045",
"language:hr",
"license:other",
"size_categories:1K<n<10K",
"task_categories:text-classification",
"tags:hate-speech-detection",
"tags:offensive-language"
] | https://huggingface.co/datasets/classla/FRENK-hate-hr/resolve/main/README.md | ---
language:
- hr
license:
- other
size_categories:
- 1K<n<10K
task_categories:
- text-classification
task_ids: []
tags:
- hate-speech-detection
- offensive-language
---
# Offensive language dataset of Croatian comments FRENK 1.0
Croatian subset of the [FRENK dataset](http://hdl.handle.net/11356/1433). Also available on HuggingFace dataset hub: [English subset](https://huggingface.co/datasets/5roop/FRENK-hate-en), [Slovenian subset](https://huggingface.co/datasets/5roop/FRENK-hate-sl).
## Dataset Description
- **Homepage:** http://hdl.handle.net/11356/1433
- **Repository:** http://hdl.handle.net/11356/1433
- **Paper:** https://arxiv.org/abs/1906.02045
- **Project page** https://nl.ijs.si/frenk/
## Description of the original dataset
>The original FRENK dataset consists of comments to Facebook posts (news articles) of mainstream media outlets from Croatia, Great Britain, and Slovenia, on the topics of migrants and LGBT. The dataset contains whole discussion threads. Each comment is annotated by the type of socially unacceptable discourse (e.g., inappropriate, offensive, violent speech) and its target (e.g., migrants/LGBT, commenters, media). The annotation schema is described in detail in [https://arxiv.org/pdf/1906.02045.pdf]. Usernames in the metadata are pseudo-anonymised and removed from the comments.
>
>The data in each language (Croatian (hr), English (en), Slovenian (sl), and topic (migrants, LGBT) is divided into a training and a testing portion. The training and testing data consist of separate discussion threads, i.e., there is no cross-discussion-thread contamination between training and testing data. The sizes of the splits are the following: Croatian, migrants: 4356 training comments, 978 testing comments; Croatian LGBT: 4494 training comments, 1142 comments; English, migrants: 4540 training comments, 1285 testing comments; English, LGBT: 4819 training comments, 1017 testing comments; Slovenian, migrants: 5145 training comments, 1277 testing comments; Slovenian, LGBT: 2842 training comments, 900 testing comments.
For this dataset only the Croatian data was used. Training segment has been split into beginning 90% (published here as training split) and end 10% (published here as dev split). Test segment has been preserved in its original form.
## Usage in `Transformers`
```python
import datasets
ds = datasets.load_dataset("classla/FRENK-hate-hr","binary")
```
For binary classification the following encoding is used:
```python
_CLASS_MAP_BINARY = {
'Acceptable': 0,
'Offensive': 1,
}
```
The original labels are available if the dataset is loaded with the `multiclass` option:
```python
import datasets
ds = datasets.load_dataset("classla/FRENK-hate-hr","multiclass").
```
In this case the encoding used is:
```python
_CLASS_MAP_MULTICLASS = {
'Acceptable speech': 0,
'Inappropriate': 1,
'Background offensive': 2,
'Other offensive': 3,
'Background violence': 4,
'Other violence': 5,
}
```
## Data structure
* `text`: text
* `target`: who is the target of the hate-speech text ("no target", "commenter", "target" (migrants or LGBT, depending on the topic), or "related to" (again, the topic))
* `topic`: whether the text relates to lgbt or migrants hate-speech domains
* `label`: label of the text instance, see above.
## Data instance
```
{'text': 'Potpisujem komentar g ankice pavicic',
'target': 'No target',
'topic': 'lgbt',
'label': 0}
```
## Licensing information
CLARIN.SI Licence ACA ID-BY-NC-INF-NORED 1.0
## Citation information
When using this dataset please cite the following paper:
```
@misc{ljubešić2019frenk,
title={The FRENK Datasets of Socially Unacceptable Discourse in Slovene and English},
author={Nikola Ljubešić and Darja Fišer and Tomaž Erjavec},
year={2019},
eprint={1906.02045},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/1906.02045}
}
```
The original dataset can be cited as
```
@misc{11356/1433,
title = {Offensive language dataset of Croatian, English and Slovenian comments {FRENK} 1.0},
author = {Ljube{\v s}i{\'c}, Nikola and Fi{\v s}er, Darja and Erjavec, Toma{\v z}},
url = {http://hdl.handle.net/11356/1433},
note = {Slovenian language resource repository {CLARIN}.{SI}},
copyright = {{CLARIN}.{SI} Licence {ACA} {ID}-{BY}-{NC}-{INF}-{NORED} 1.0},
year = {2021} }
``` |
classla | null | @misc{ljubešić2019frenk,
title={The FRENK Datasets of Socially Unacceptable Discourse in Slovene and English},
author={Nikola Ljubešić and Darja Fišer and Tomaž Erjavec},
year={2019},
eprint={1906.02045},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/1906.02045}
} | The FRENK Datasets of Socially Unacceptable Discourse in Slovene. | false | 482 | false | classla/FRENK-hate-sl | 2022-10-21T07:46:11.000Z | null | false | 37c8b42c63d4eb75f549679158a85eb5bd984caa | [] | [
"arxiv:1906.02045",
"language:sl",
"license:other",
"size_categories:1K<n<10K",
"task_categories:text-classification",
"tags:hate-speech-detection",
"tags:offensive-language"
] | https://huggingface.co/datasets/classla/FRENK-hate-sl/resolve/main/README.md | ---
language:
- sl
license:
- other
size_categories:
- 1K<n<10K
task_categories:
- text-classification
task_ids: []
tags:
- hate-speech-detection
- offensive-language
---
Slovenian subset of the [FRENK dataset](http://hdl.handle.net/11356/1433). Also available on HuggingFace dataset hub: [English subset](https://huggingface.co/datasets/5roop/FRENK-hate-en), [Croatian subset](https://huggingface.co/datasets/5roop/FRENK-hate-hr).
## Dataset Description
- **Homepage:** http://hdl.handle.net/11356/1433
- **Repository:** http://hdl.handle.net/11356/1433
- **Paper:** https://arxiv.org/abs/1906.02045
- **Project page** https://nl.ijs.si/frenk/
## Description of the original dataset
>The original FRENK dataset consists of comments to Facebook posts (news articles) of mainstream media outlets from Croatia, Great Britain, and Slovenia, on the topics of migrants and LGBT. The dataset contains whole discussion threads. Each comment is annotated by the type of socially unacceptable discourse (e.g., inappropriate, offensive, violent speech) and its target (e.g., migrants/LGBT, commenters, media). The annotation schema is described in detail in [https://arxiv.org/pdf/1906.02045.pdf]. Usernames in the metadata are pseudo-anonymised and removed from the comments.
>
>The data in each language (Croatian (hr), English (en), Slovenian (sl), and topic (migrants, LGBT) is divided into a training and a testing portion. The training and testing data consist of separate discussion threads, i.e., there is no cross-discussion-thread contamination between training and testing data. The sizes of the splits are the following: Croatian, migrants: 4356 training comments, 978 testing comments; Croatian LGBT: 4494 training comments, 1142 comments; English, migrants: 4540 training comments, 1285 testing comments; English, LGBT: 4819 training comments, 1017 testing comments; Slovenian, migrants: 5145 training comments, 1277 testing comments; Slovenian, LGBT: 2842 training comments, 900 testing comments.
For this dataset only the Croatian data was used. Training segment has been split into beginning 90% (published here as training split) and end 10% (published here as dev split).
## Usage in `Transformers`
```python
import datasets
ds = datasets.load_dataset("classla/FRENK-hate-sl","binary")
```
For binary classification the following encoding is used:
```python
_CLASS_MAP_BINARY = {
'Acceptable': 0,
'Offensive': 1,
}
```
The original labels are available if the dataset is loaded with the `multiclass` option:
```python
import datasets
ds = datasets.load_dataset("classla/FRENK-hate-sl","multiclass").
```
In this case the encoding used is:
```python
_CLASS_MAP_MULTICLASS = {
'Acceptable speech': 0,
'Inappropriate': 1,
'Background offensive': 2,
'Other offensive': 3,
'Background violence': 4,
'Other violence': 5,
}
```
## Data structure
* `text`: text
* `target`: who is the target of the hate-speech text ("no target", "commenter", "target" (migrants or LGBT, depending on the topic), or "related to" (again, the topic))
* `topic`: whether the text relates to lgbt or migrants hate-speech domains
* `label`: label of the text instance, see above.
## Data instance
```
{'text': 'Otroci so odprti in brez predsodkov.Predsodke jim vcepimo starejši,starši,družba,družina...Če otroku lepo razložimo,razume.Nikoli ni dobro,da omejujemo otroka,njegovo inteligenco in duhovnost z lastnim ne razumevanjem nečesa ali nekoga.Predsodek je miselni zapor,prepreka,da bi bili svobodni.Ljubezen je svoboda.Sem ZA spremembo zakona!Srečno :D',
'target': 'No target',
'topic': 'lgbt',
'label': 0}
```
## Licensing information
CLARIN.SI Licence ACA ID-BY-NC-INF-NORED 1.0
## Citation information
When using this dataset please cite the following paper:
```
@misc{ljubešić2019frenk,
title={The FRENK Datasets of Socially Unacceptable Discourse in Slovene and English},
author={Nikola Ljubešić and Darja Fišer and Tomaž Erjavec},
year={2019},
eprint={1906.02045},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/1906.02045}
}
```
The original dataset can be cited as
```
@misc{11356/1433,
title = {Offensive language dataset of Croatian, English and Slovenian comments {FRENK} 1.0},
author = {Ljube{\v s}i{\'c}, Nikola and Fi{\v s}er, Darja and Erjavec, Toma{\v z}},
url = {http://hdl.handle.net/11356/1433},
note = {Slovenian language resource repository {CLARIN}.{SI}},
copyright = {{CLARIN}.{SI} Licence {ACA} {ID}-{BY}-{NC}-{INF}-{NORED} 1.0},
year = {2021} }
```
|
classla | null | @article{DBLP:journals/corr/abs-2104-09243,
author = {Nikola Ljubesic and
Davor Lauc},
title = {BERTi{\'{c}} - The Transformer Language Model for Bosnian, Croatian,
Montenegrin and Serbian},
journal = {CoRR},
volume = {abs/2104.09243},
year = {2021},
url = {https://arxiv.org/abs/2104.09243},
archivePrefix = {arXiv},
} | The COPA-HR dataset (Choice of plausible alternatives in Croatian) is a translation
of the English COPA dataset (https://people.ict.usc.edu/~gordon/copa.html) by following the
XCOPA dataset translation methodology (https://arxiv.org/abs/2005.00333). The dataset consists of 1000 premises
(My body cast a shadow over the grass), each given a question (What is the cause?), and two choices
(The sun was rising; The grass was cut), with a label encoding which of the choices is more plausible
given the annotator or translator (The sun was rising).
The dataset is split into 400 training samples, 100 validation samples, and 500 test samples. It includes the
following features: 'premise', 'choice1', 'choice2', 'label', 'question', 'changed' (boolean). | false | 681 | false | classla/copa_hr | 2022-10-25T07:32:15.000Z | null | false | f3f3a4708e6f8b92915ab02c20ac7fb829e45173 | [] | [
"arxiv:2005.00333",
"arxiv:2104.09243",
"language:hr",
"license:cc-by-sa-4.0",
"task_categories:text-classification",
"task_ids:natural-language-inference",
"tags:causal-reasoning",
"tags:textual-entailment",
"tags:commonsense-reasoning"
] | https://huggingface.co/datasets/classla/copa_hr/resolve/main/README.md | ---
language:
- hr
license:
- cc-by-sa-4.0
task_categories:
- text-classification
task_ids:
- natural-language-inference
tags:
- causal-reasoning
- textual-entailment
- commonsense-reasoning
---
The COPA-HR dataset (Choice of plausible alternatives in Croatian) is a translation
of the English COPA dataset (https://people.ict.usc.edu/~gordon/copa.html) by following the
XCOPA dataset translation methodology (https://arxiv.org/abs/2005.00333). The dataset consists of 1000 premises
(My body cast a shadow over the grass), each given a question (What is the cause?), and two choices
(The sun was rising; The grass was cut), with a label encoding which of the choices is more plausible
given the annotator or translator (The sun was rising).
The dataset is split into 400 training samples, 100 validation samples, and 500 test samples. It includes the
following features: 'premise', 'choice1', 'choice2', 'label', 'question', 'changed' (boolean).
If you use the dataset in your work, please cite
```
@article{DBLP:journals/corr/abs-2104-09243,
author = {Nikola Ljube\\\\v{s}i\\\\'{c} and
Davor Lauc},
title = {BERTi{\\\\'{c}} - The Transformer Language Model for Bosnian, Croatian,
Montenegrin and Serbian},
journal = {CoRR},
volume = {abs/2104.09243},
year = {2021},
url = {https://arxiv.org/abs/2104.09243},
archivePrefix = {arXiv},
}
``` |
classla | null | null | The hr500k training corpus contains about 500,000 tokens manually annotated on the levels of
tokenisation, sentence segmentation, morphosyntactic tagging, lemmatisation and named entities.
On the sentence level, the dataset contains 20159 training samples, 1963 validation samples and 2672 test samples
across the respective data splits. Each sample represents a sentence and includes the following features:
sentence ID ('sent_id'), sentence text ('text'), list of tokens ('tokens'), list of lemmas ('lemmas'),
list of Multext-East tags ('xpos_tags), list of UPOS tags ('upos_tags'),
list of morphological features ('feats'), and list of IOB tags ('iob_tags'). The 'upos_tags' and 'iob_tags' features
are encoded as class labels. | false | 637 | false | classla/hr500k | 2022-10-25T07:32:05.000Z | null | false | 708662e326e2e0ee4ce0fb7fa4e41db6c93771f0 | [] | [
"language:hr",
"license:cc-by-sa-4.0",
"task_categories:other",
"task_ids:lemmatization",
"task_ids:named-entity-recognition",
"task_ids:part-of-speech",
"tags:structure-prediction",
"tags:normalization",
"tags:tokenization"
] | https://huggingface.co/datasets/classla/hr500k/resolve/main/README.md | ---
language:
- hr
license:
- cc-by-sa-4.0
task_categories:
- other
task_ids:
- lemmatization
- named-entity-recognition
- part-of-speech
tags:
- structure-prediction
- normalization
- tokenization
---
The hr500k training corpus contains 506,457 Croatian tokens manually annotated on the levels of tokenisation, sentence segmentation, morphosyntactic tagging, lemmatisation, named entities and dependency syntax.
On the sentence level, the dataset contains 20159 training samples, 1963 validation samples and 2672 test samples
across the respective data splits. Each sample represents a sentence and includes the following features:
sentence ID ('sent\_id'), sentence text ('text'), list of tokens ('tokens'), list of lemmas ('lemmas'),
list of MULTEXT-East tags ('xpos\_tags), list of UPOS tags ('upos\_tags'), list of morphological features ('feats'),
and list of IOB tags ('iob\_tags'). A subset of the data also contains universal dependencies ('ud') and consists of
7498 training samples, 649 validation samples, and 742 test samples.
Three dataset configurations are available, namely 'ner', 'upos', and 'ud', with the corresponding features
encoded as class labels. If the configuration is not specified, it defaults to 'ner'.
If you use this dataset in your research, please cite the following paper:
```
Bibtex @InProceedings{LJUBEI16.340,
author = {Nikola Ljubešić and Filip Klubička and Željko Agić and Ivo-Pavao Jazbec},
title = {New Inflectional Lexicons and Training Corpora for Improved Morphosyntactic Annotation of Croatian and Serbian},
booktitle = {Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC 2016)},
year = {2016},
month = {may},
date = {23-28},
location = {Portorož, Slovenia},
editor = {Nicoletta Calzolari (Conference Chair) and Khalid Choukri and Thierry Declerck and Sara Goggi and Marko Grobelnik and Bente Maegaard and Joseph Mariani and Helene Mazo and Asuncion Moreno and Jan Odijk and Stelios Piperidis},
publisher = {European Language Resources Association (ELRA)},
address = {Paris, France},
isbn = {978-2-9517408-9-1},
language = {english}
}
``` |
classla | null | null | The dataset contains 6273 training samples, 762 validation samples and 749 test samples.
Each sample represents a sentence and includes the following features: sentence ID ('sent_id'),
list of tokens ('tokens'), list of normalised word forms ('norms'), list of lemmas ('lemmas'),
list of Multext-East tags ('xpos_tags), list of morphological features ('feats'),
and list of UPOS tags ('upos_tags'), which are encoded as class labels. | false | 323 | false | classla/janes_tag | 2022-10-25T07:31:04.000Z | null | false | ba014295e666710c5dfe6215338933ecf235156c | [] | [
"language:si",
"license:cc-by-sa-4.0",
"task_categories:other",
"task_ids:lemmatization",
"task_ids:part-of-speech",
"tags:structure-prediction",
"tags:normalization",
"tags:tokenization"
] | https://huggingface.co/datasets/classla/janes_tag/resolve/main/README.md | ---
language:
- si
license:
- cc-by-sa-4.0
task_categories:
- other
task_ids:
- lemmatization
- part-of-speech
tags:
- structure-prediction
- normalization
- tokenization
---
The dataset contains 6273 training samples, 762 validation samples and 749 test samples.
Each sample represents a sentence and includes the following features: sentence ID ('sent\_id'),
list of tokens ('tokens'), list of normalised word forms ('norms'), list of lemmas ('lemmas'),
list of Multext-East tags ('xpos\_tags), list of morphological features ('feats'),
and list of UPOS tags ('upos\_tags'), which are encoded as class labels. |
classla | null | null | The dataset contains 6339 training samples, 815 validation samples and 785 test samples.
Each sample represents a sentence and includes the following features: sentence ID ('sent_id'),
list of tokens ('tokens'), list of lemmas ('lemmas'), list of UPOS tags ('upos_tags'),
list of Multext-East tags ('xpos_tags), list of morphological features ('feats'),
and list of IOB tags ('iob_tags'), which are encoded as class labels. | false | 324 | false | classla/reldi_hr | 2022-10-25T07:30:56.000Z | null | false | da293b9a70a87a936777e93dd59046ddbc6399ce | [] | [
"language:hr",
"license:cc-by-sa-4.0",
"task_categories:other",
"task_ids:lemmatization",
"task_ids:named-entity-recognition",
"task_ids:part-of-speech",
"tags:structure-prediction",
"tags:normalization",
"tags:tokenization"
] | https://huggingface.co/datasets/classla/reldi_hr/resolve/main/README.md | ---
language:
- hr
license:
- cc-by-sa-4.0
task_categories:
- other
task_ids:
- lemmatization
- named-entity-recognition
- part-of-speech
tags:
- structure-prediction
- normalization
- tokenization
---
This dataset is based on 3,871 Croatian tweets that were segmented into sentences, tokens, and annotated with normalized forms, lemmas, MULTEXT-East tags (XPOS), UPOS tags and morphological features, and named entities.
The dataset contains 6339 training samples (sentences), 815 validation samples and 785 test samples.
Each sample represents a sentence and includes the following features: sentence ID ('sent\_id'),
list of tokens ('tokens'), list of normalised tokens ('norms'), list of lemmas ('lemmas'), list of UPOS tags ('upos\_tags'),
list of MULTEXT-East tags ('xpos\_tags), list of morphological features ('feats'),
and list of named entity IOB tags ('iob\_tags'), which are encoded as class labels.
If you are using this dataset in your research, please cite the following paper:
```
@article{Miličević_Ljubešić_2016,
title={Tviterasi, tviteraši or twitteraši? Producing and analysing a normalised dataset of Croatian and Serbian tweets},
volume={4},
url={https://revije.ff.uni-lj.si/slovenscina2/article/view/7007},
DOI={10.4312/slo2.0.2016.2.156-188},
number={2},
journal={Slovenščina 2.0: empirical, applied and interdisciplinary research},
author={Miličević, Maja and Ljubešić, Nikola},
year={2016},
month={Sep.},
pages={156–188} }
```
|
classla | null | null | The dataset contains 5462 training samples, 711 validation samples and 725 test samples.
Each sample represents a sentence and includes the following features: sentence ID ('sent_id'),
list of tokens ('tokens'), list of lemmas ('lemmas'), list of UPOS tags ('upos_tags'),
list of Multext-East tags ('xpos_tags), list of morphological features ('feats'),
and list of IOB tags ('iob_tags'), which are encoded as class labels. | false | 323 | false | classla/reldi_sr | 2022-10-25T07:30:33.000Z | null | false | 10a37a1a9ea782093646e0b03d5ef05b3e1e11d5 | [] | [
"language:sr",
"license:cc-by-sa-4.0",
"task_categories:other",
"task_ids:lemmatization",
"task_ids:named-entity-recognition",
"task_ids:part-of-speech",
"tags:structure-prediction",
"tags:normalization",
"tags:tokenization"
] | https://huggingface.co/datasets/classla/reldi_sr/resolve/main/README.md | ---
language:
- sr
license:
- cc-by-sa-4.0
task_categories:
- other
task_ids:
- lemmatization
- named-entity-recognition
- part-of-speech
tags:
- structure-prediction
- normalization
- tokenization
---
This dataset is based on 3,748 Serbian tweets that were segmented into sentences, tokens, and annotated with normalized forms, lemmas, MULTEXT-East tags (XPOS), UPOS tags and morphological features, and named entities.
The dataset contains 5462 training samples (sentences), 711 validation samples and 725 test samples.
Each sample represents a sentence and includes the following features: sentence ID ('sent\_id'),
list of tokens ('tokens'), list of normalised tokens ('norms'), list of lemmas ('lemmas'), list of UPOS tags ('upos\_tags'),
list of MULTEXT-East tags ('xpos\_tags), list of morphological features ('feats'),
and list of named entity IOB tags ('iob\_tags'), which are encoded as class labels.
If you are using this dataset in your research, please cite the following paper:
```
@article{Miličević_Ljubešić_2016,
title={Tviterasi, tviteraši or twitteraši? Producing and analysing a normalised dataset of Croatian and Serbian tweets},
volume={4},
url={https://revije.ff.uni-lj.si/slovenscina2/article/view/7007},
DOI={10.4312/slo2.0.2016.2.156-188},
number={2},
journal={Slovenščina 2.0: empirical, applied and interdisciplinary research},
author={Miličević, Maja and Ljubešić, Nikola},
year={2016},
month={Sep.},
pages={156–188} }
``` |
classla | null | null | SETimes_sr is a Serbian dataset annotated for morphosyntactic information and named entities.
The dataset contains 3177 training samples, 395 validation samples and 319 test samples
across the respective data splits. Each sample represents a sentence and includes the following features:
sentence ID ('sent_id'), sentence text ('text'), list of tokens ('tokens'), list of lemmas ('lemmas'),
list of Multext-East tags ('xpos_tags), list of UPOS tags ('upos_tags'),
list of morphological features ('feats'), and list of IOB tags ('iob_tags'). The 'upos_tags' and 'iob_tags' features
are encoded as class labels. | false | 637 | false | classla/setimes_sr | 2022-10-25T07:30:04.000Z | null | false | 42861d4054bc5fb993e6606e3c70a2957ec52e91 | [] | [
"language:sr",
"license:cc-by-sa-4.0",
"task_categories:other",
"task_ids:lemmatization",
"task_ids:named-entity-recognition",
"task_ids:part-of-speech",
"tags:structure-prediction",
"tags:normalization",
"tags:tokenization"
] | https://huggingface.co/datasets/classla/setimes_sr/resolve/main/README.md | ---
language:
- sr
license:
- cc-by-sa-4.0
task_categories:
- other
task_ids:
- lemmatization
- named-entity-recognition
- part-of-speech
tags:
- structure-prediction
- normalization
- tokenization
---
The SETimes\_sr training corpus contains 86,726 Serbian tokens manually annotated on the levels of tokenisation, sentence segmentation, morphosyntactic tagging, lemmatisation, named entities and dependency syntax.
The dataset contains 3177 training samples, 395 validation samples and 319 test samples
across the respective data splits. Each sample represents a sentence and includes the following features:
sentence ID ('sent\_id'), sentence text ('text'), list of tokens ('tokens'), list of lemmas ('lemmas'),
list of MULTEXT-East tags ('xpos\_tags), list of UPOS tags ('upos\_tags'),
list of morphological features ('feats'), list of IOB tags ('iob\_tags') and list of universal dependencies ('uds').
Three dataset configurations are available, namely 'ner', 'upos', and 'ud', with the corresponding features
encoded as class labels. If the configuration is not specified, it defaults to 'ner'.
If you use this dataset in your research, please cite the following paper:
```
@inproceedings{samardzic-etal-2017-universal,
title = "{U}niversal {D}ependencies for {S}erbian in Comparison with {C}roatian and Other {S}lavic Languages",
author = "Samard{\v{z}}i{\'c}, Tanja and
Starovi{\'c}, Mirjana and
Agi{\'c}, {\v{Z}}eljko and
Ljube{\v{s}}i{\'c}, Nikola",
booktitle = "Proceedings of the 6th Workshop on {B}alto-{S}lavic Natural Language Processing",
month = apr,
year = "2017",
address = "Valencia, Spain",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/W17-1407",
doi = "10.18653/v1/W17-1407",
pages = "39--44",
}
``` |
classla | null | null | The dataset contains 7432 training samples, 1164 validation samples and 893 test samples.
Each sample represents a sentence and includes the following features: sentence ID ('sent_id'),
list of tokens ('tokens'), list of lemmas ('lemmas'),
list of Multext-East tags ('xpos_tags), list of UPOS tags ('upos_tags'), list of morphological features ('feats'),
list of IOB tags ('iob_tags'), and list of universal dependency tags ('uds'). Three dataset configurations are
available, where the corresponding features are encoded as class labels: 'ner', 'upos', and 'ud'. | false | 637 | false | classla/ssj500k | 2022-10-28T05:37:22.000Z | null | false | 446b04c97cb43772a229cebbb8da0ce05ee03d2d | [] | [
"language:sl",
"license:cc-by-sa-4.0",
"task_categories:token-classification",
"task_ids:lemmatization",
"task_ids:named-entity-recognition",
"task_ids:parsing",
"task_ids:part-of-speech",
"tags:structure-prediction",
"tags:tokenization",
"tags:dependency-parsing"
] | https://huggingface.co/datasets/classla/ssj500k/resolve/main/README.md | ---
language:
- sl
license:
- cc-by-sa-4.0
task_categories:
- token-classification
task_ids:
- lemmatization
- named-entity-recognition
- parsing
- part-of-speech
tags:
- structure-prediction
- tokenization
- dependency-parsing
---
The dataset contains 7432 training samples, 1164 validation samples and 893 test samples.
Each sample represents a sentence and includes the following features: sentence ID ('sent\_id'),
list of tokens ('tokens'), list of lemmas ('lemmas'),
list of Multext-East tags ('xpos\_tags), list of UPOS tags ('upos\_tags'), list of morphological features ('feats'),
list of IOB tags ('iob\_tags'), and list of universal dependency tags ('uds'). Three dataset configurations are
available, where the corresponding features are encoded as class labels: 'ner', 'upos', and 'ud'. |
clem | null | null | null | false | 166 | false | clem/autonlp-data-french_word_detection | 2021-09-14T09:45:38.000Z | null | false | dcbb0c37d501225a976dc9e8a12bf0e20c8e2e04 | [] | [] | https://huggingface.co/datasets/clem/autonlp-data-french_word_detection/resolve/main/README.md | This is a very good dataset! |
clips | null | @InProceedings{mfaq_a_multilingual_dataset,
title={MFAQ: a Multilingual FAQ Dataset},
author={Maxime {De Bruyn} and Ehsan Lotfi and Jeska Buhmann and Walter Daelemans},
year={2021},
booktitle={MRQA @ EMNLP 2021}
} | We present the first multilingual FAQ dataset publicly available. We collected around 6M FAQ pairs from the web, in 21 different languages. | false | 7,127 | false | clips/mfaq | 2022-10-20T11:32:50.000Z | null | false | 87a7bada8da4fe2a7b738c6d3e549153383198ad | [] | [
"arxiv:2109.12870",
"annotations_creators:no-annotation",
"language_creators:other",
"language:cs",
"language:da",
"language:de",
"language:en",
"language:es",
"language:fi",
"language:fr",
"language:he",
"language:hr",
"language:hu",
"language:id",
"language:it",
"language:nl",
"language:no",
"language:pl",
"language:pt",
"language:ro",
"language:ru",
"language:sv",
"language:tr",
"language:vi",
"license:cc0-1.0",
"multilinguality:multilingual",
"size_categories:unknown",
"source_datasets:original",
"task_categories:question-answering",
"task_ids:multiple-choice-qa"
] | https://huggingface.co/datasets/clips/mfaq/resolve/main/README.md | ---
annotations_creators:
- no-annotation
language_creators:
- other
language:
- cs
- da
- de
- en
- es
- fi
- fr
- he
- hr
- hu
- id
- it
- nl
- 'no'
- pl
- pt
- ro
- ru
- sv
- tr
- vi
license:
- cc0-1.0
multilinguality:
- multilingual
pretty_name: MFAQ - a Multilingual FAQ Dataset
size_categories:
- unknown
source_datasets:
- original
task_categories:
- question-answering
task_ids:
- multiple-choice-qa
---
# MFAQ
🚨 See [MQA](https://huggingface.co/datasets/clips/mqa) or [MFAQ Light](maximedb/mfaq_light) for an updated version of the dataset.
MFAQ is a multilingual corpus of *Frequently Asked Questions* parsed from the [Common Crawl](https://commoncrawl.org/).
```
from datasets import load_dataset
load_dataset("clips/mfaq", "en")
{
"qa_pairs": [
{
"question": "Do I need a rental Car in Cork?",
"answer": "If you plan on travelling outside of Cork City, for instance to Kinsale [...]"
},
...
]
}
```
## Languages
We collected around 6M pairs of questions and answers in 21 different languages. To download a language specific subset you need to specify the language key as configuration. See below for an example.
```
load_dataset("clips/mfaq", "en") # replace "en" by any language listed below
```
| Language | Key | Pairs | Pages |
|------------|-----|-----------|-----------|
| All | all | 6,346,693 | 1,035,649 |
| English | en | 3,719,484 | 608,796 |
| German | de | 829,098 | 111,618 |
| Spanish | es | 482,818 | 75,489 |
| French | fr | 351,458 | 56,317 |
| Italian | it | 155,296 | 24,562 |
| Dutch | nl | 150,819 | 32,574 |
| Portuguese | pt | 138,778 | 26,169 |
| Turkish | tr | 102,373 | 19,002 |
| Russian | ru | 91,771 | 22,643 |
| Polish | pl | 65,182 | 10,695 |
| Indonesian | id | 45,839 | 7,910 |
| Norwegian | no | 37,711 | 5,143 |
| Swedish | sv | 37,003 | 5,270 |
| Danish | da | 32,655 | 5,279 |
| Vietnamese | vi | 27,157 | 5,261 |
| Finnish | fi | 20,485 | 2,795 |
| Romanian | ro | 17,066 | 3,554 |
| Czech | cs | 16,675 | 2,568 |
| Hebrew | he | 11,212 | 1,921 |
| Hungarian | hu | 8,598 | 1,264 |
| Croatian | hr | 5,215 | 819 |
## Data Fields
#### Nested (per page - default)
The data is organized by page. Each page contains a list of questions and answers.
- **id**
- **language**
- **num_pairs**: the number of FAQs on the page
- **domain**: source web domain of the FAQs
- **qa_pairs**: a list of questions and answers
- **question**
- **answer**
- **language**
#### Flattened
The data is organized by pair (i.e. pages are flattened). You can access the flat version of any language by appending `_flat` to the configuration (e.g. `en_flat`). The data will be returned pair-by-pair instead of page-by-page.
- **domain_id**
- **pair_id**
- **language**
- **domain**: source web domain of the FAQs
- **question**
- **answer**
## Source Data
This section was adapted from the source data description of [OSCAR](https://huggingface.co/datasets/oscar#source-data)
Common Crawl is a non-profit foundation which produces and maintains an open repository of web crawled data that is both accessible and analysable. Common Crawl's complete web archive consists of petabytes of data collected over 8 years of web crawling. The repository contains raw web page HTML data (WARC files), metdata extracts (WAT files) and plain text extracts (WET files). The organisation's crawlers has always respected nofollow and robots.txt policies.
To construct MFAQ, the WARC files of Common Crawl were used. We looked for `FAQPage` markup in the HTML and subsequently parsed the `FAQItem` from the page.
## People
This model was developed by [Maxime De Bruyn](https://www.linkedin.com/in/maximedebruyn/), Ehsan Lotfi, Jeska Buhmann and Walter Daelemans.
## Licensing Information
```
These data are released under this licensing scheme.
We do not own any of the text from which these data has been extracted.
We license the actual packaging of these data under the Creative Commons CC0 license ("no rights reserved") http://creativecommons.org/publicdomain/zero/1.0/
Should you consider that our data contains material that is owned by you and should therefore not be reproduced here, please:
* Clearly identify yourself, with detailed contact data such as an address, telephone number or email address at which you can be contacted.
* Clearly identify the copyrighted work claimed to be infringed.
* Clearly identify the material that is claimed to be infringing and information reasonably sufficient to allow us to locate the material.
We will comply to legitimate requests by removing the affected sources from the next release of the corpus.
```
## Citation information
```
@misc{debruyn2021mfaq,
title={MFAQ: a Multilingual FAQ Dataset},
author={Maxime {De Bruyn} and Ehsan Lotfi and Jeska Buhmann and Walter Daelemans},
year={2021},
eprint={2109.12870},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` |
clips | null | @misc{debruyn2021mfaq,
title={MFAQ: a Multilingual FAQ Dataset},
author={Maxime {De Bruyn} and Ehsan Lotfi and Jeska Buhmann and Walter Daelemans},
year={2021},
booktitle={MRQA@EMNLP2021},
} | MQA is a multilingual corpus of questions and answers parsed from the Common Crawl. Questions are divided between Frequently Asked Questions (FAQ) pages and Community Question Answering (CQA) pages. | false | 42,844 | false | clips/mqa | 2022-09-27T12:38:50.000Z | null | false | 27eebc4a00d229f8dd4ae2a6d9f1e4ad45781f3b | [] | [
"annotations_creators:no-annotation",
"language_creators:other",
"language:ca",
"language:en",
"language:de",
"language:es",
"language:fr",
"language:ru",
"language:ja",
"language:it",
"language:zh",
"language:pt",
"language:nl",
"language:tr",
"language:pl",
"language:vi",
"language:ar",
"language:id",
"language:uk",
"language:ro",
"language:no",
"language:th",
"language:sv",
"language:el",
"language:fi",
"language:he",
"language:da",
"language:cs",
"language:ko",
"language:fa",
"language:hi",
"language:hu",
"language:sk",
"language:lt",
"language:et",
"language:hr",
"language:is",
"language:lv",
"language:ms",
"language:bg",
"language:sr",
"license:cc0-1.0",
"multilinguality:multilingual",
"size_categories:unknown",
"source_datasets:original",
"task_categories:question-answering",
"task_ids:multiple-choice-qa"
] | https://huggingface.co/datasets/clips/mqa/resolve/main/README.md | ---
annotations_creators:
- no-annotation
language_creators:
- other
language:
- ca
- en
- de
- es
- fr
- ru
- ja
- it
- zh
- pt
- nl
- tr
- pl
- vi
- ar
- id
- uk
- ro
- no
- th
- sv
- el
- fi
- he
- da
- cs
- ko
- fa
- hi
- hu
- sk
- lt
- et
- hr
- is
- lv
- ms
- bg
- sr
- ca
license:
- cc0-1.0
multilinguality:
- multilingual
pretty_name: MQA - a Multilingual FAQ and CQA Dataset
size_categories:
- unknown
source_datasets:
- original
task_categories:
- question-answering
task_ids:
- multiple-choice-qa
---
# MQA
MQA is a Multilingual corpus of Questions and Answers (MQA) parsed from the [Common Crawl](https://commoncrawl.org/). Questions are divided in two types: *Frequently Asked Questions (FAQ)* and *Community Question Answering (CQA)*.
```python
from datasets import load_dataset
all_data = load_dataset("clips/mqa", language="en")
{
"name": "the title of the question (if any)",
"text": "the body of the question (if any)",
"answers": [{
"text": "the text of the answer",
"is_accepted": "true|false"
}]
}
faq_data = load_dataset("clips/mqa", scope="faq", language="en")
cqa_data = load_dataset("clips/mqa", scope="cqa", language="en")
```
## Languages
We collected around **234M pairs** of questions and answers in **39 languages**. To download a language specific subset you need to specify the language key as configuration. See below for an example.
```python
load_dataset("clips/mqa", language="en") # replace "en" by any language listed below
```
| Language | FAQ | CQA |
|:-----------|------------:|-----------:|
| en | 174,696,414 | 14,082,180 |
| de | 17,796,992 | 1,094,606 |
| es | 14,967,582 | 845,836 |
| fr | 13,096,727 | 1,299,359 |
| ru | 12,435,022 | 1,715,131 |
| it | 6,850,573 | 455,027 |
| ja | 6,369,706 | 2,089,952 |
| zh | 5,940,796 | 579,596 |
| pt | 5,851,286 | 373,982 |
| nl | 4,882,511 | 503,376 |
| tr | 3,893,964 | 370,975 |
| pl | 3,766,531 | 70,559 |
| vi | 2,795,227 | 96,528 |
| id | 2,253,070 | 200,441 |
| ar | 2,211,795 | 805,661 |
| uk | 2,090,611 | 27,260 |
| el | 1,758,618 | 17,167 |
| no | 1,752,820 | 11,786 |
| sv | 1,733,582 | 20,024 |
| fi | 1,717,221 | 41,371 |
| ro | 1,689,471 | 93,222 |
| th | 1,685,463 | 73,204 |
| da | 1,554,581 | 16,398 |
| he | 1,422,449 | 88,435 |
| ko | 1,361,901 | 49,061 |
| cs | 1,224,312 | 143,863 |
| hu | 878,385 | 27,639 |
| fa | 787,420 | 118,805 |
| sk | 785,101 | 4,615 |
| lt | 672,105 | 301 |
| et | 547,208 | 441 |
| hi | 516,342 | 205,645 |
| hr | 458,958 | 11,677 |
| is | 437,748 | 37 |
| lv | 428,002 | 88 |
| ms | 230,568 | 7,460 |
| bg | 198,671 | 5,320 |
| sr | 110,270 | 3,980 |
| ca | 100,201 | 1,914 |
## FAQ vs. CQA
You can download the *Frequently Asked Questions* (FAQ) or the *Community Question Answering* (CQA) part of the dataset.
```python
faq = load_dataset("clips/mqa", scope="faq")
cqa = load_dataset("clips/mqa", scope="cqa")
all = load_dataset("clips/mqa", scope="all")
```
Although FAQ and CQA questions share the same structure, CQA questions can have multiple answers for a given questions, while FAQ questions have a single answer. FAQ questions typically only have a title (`name` key), while CQA have a title and a body (`name` and `text`).
## Nesting and Data Fields
You can specify three different nesting level: `question`, `page` and `domain`.
#### Question
```python
load_dataset("clips/mqa", level="question") # default
```
The default level is the question object:
- **name**: the title of the question(if any) in markdown format
- **text**: the body of the question (if any) in markdown format
- **answers**: a list of answers
- **text**: the title of the answer (if any) in markdown format
- **name**: the body of the answer in markdown format
- **is_accepted**: true if the answer is selected.
#### Page
This level returns a list of questions present on the same page. This is mostly useful for FAQs since CQAs already have one question per page.
```python
load_dataset("clips/mqa", level="page")
```
#### Domain
This level returns a list of pages present on the web domain. This is a good way to cope with FAQs duplication by sampling one page per domain at each epoch.
```python
load_dataset("clips/mqa", level="domain")
```
## Source Data
This section was adapted from the source data description of [OSCAR](https://huggingface.co/datasets/oscar#source-data)
Common Crawl is a non-profit foundation which produces and maintains an open repository of web crawled data that is both accessible and analysable. Common Crawl's complete web archive consists of petabytes of data collected over 8 years of web crawling. The repository contains raw web page HTML data (WARC files), metdata extracts (WAT files) and plain text extracts (WET files). The organisation's crawlers has always respected nofollow and robots.txt policies.
To construct MQA, we used the WARC files of Common Crawl.
## People
This model was developed by [Maxime De Bruyn](https://maximedb.vercel.app), Ehsan Lotfi, Jeska Buhmann and Walter Daelemans.
## Licensing Information
```
These data are released under this licensing scheme.
We do not own any of the text from which these data has been extracted.
We license the actual packaging of these data under the Creative Commons CC0 license ("no rights reserved") http://creativecommons.org/publicdomain/zero/1.0/
Should you consider that our data contains material that is owned by you and should therefore not be reproduced here, please:
* Clearly identify yourself, with detailed contact data such as an address, telephone number or email address at which you can be contacted.
* Clearly identify the copyrighted work claimed to be infringed.
* Clearly identify the material that is claimed to be infringing and information reasonably sufficient to allow us to locate the material.
We will comply to legitimate requests by removing the affected sources from the next release of the corpus.
```
## Citation information
```
@inproceedings{de-bruyn-etal-2021-mfaq,
title = "{MFAQ}: a Multilingual {FAQ} Dataset",
author = "De Bruyn, Maxime and
Lotfi, Ehsan and
Buhmann, Jeska and
Daelemans, Walter",
booktitle = "Proceedings of the 3rd Workshop on Machine Reading for Question Answering",
month = nov,
year = "2021",
address = "Punta Cana, Dominican Republic",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.mrqa-1.1",
pages = "1--13",
}
``` |
cnrcastroli | null | null | null | false | 164 | false | cnrcastroli/aaaa | 2021-03-04T21:51:21.000Z | null | false | 3a1dc9acf1e9957e628865fa9937a70f71cf5f3f | [] | [] | https://huggingface.co/datasets/cnrcastroli/aaaa/resolve/main/README.md | fwefwefewf |
coastalcph | null | @inproceedings{chalkidis-etal-2022-fairlex,
author={Chalkidis, Ilias and Passini, Tommaso and Zhang, Sheng and
Tomada, Letizia and Schwemer, Sebastian Felix and Søgaard, Anders},
title={FairLex: A Multilingual Benchmark for Evaluating Fairness in Legal Text Processing},
booktitle={Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics},
year={2022},
address={Dublin, Ireland}
} | Fairlex: A multilingual benchmark for evaluating fairness in legal text processing. | false | 142 | false | coastalcph/fairlex | 2022-10-20T19:44:27.000Z | null | false | 9c674e0bf7afe89ab6e6de354594081955248a05 | [] | [
"arxiv:2103.13868",
"arxiv:2105.03887",
"annotations_creators:found",
"annotations_creators:machine-generated",
"language_creators:found",
"language:en",
"language:de",
"language:fr",
"language:it",
"language:zh",
"license:cc-by-nc-sa-4.0",
"multilinguality:monolingual",
"multilinguality:multilingual",
"size_categories:10K<n<100K",
"size_categories:1K<n<10K",
"size_categories:100K<n<1M",
"source_datasets:extended",
"task_categories:text-classification",
"task_ids:multi-label-classification",
"task_ids:multi-class-classification",
"task_ids:topic-classification",
"tags:bias",
"tags:gender-bias"
] | https://huggingface.co/datasets/coastalcph/fairlex/resolve/main/README.md | ---
annotations_creators:
- found
- machine-generated
language_creators:
- found
language:
- en
- en
- de
- fr
- it
- zh
license:
- cc-by-nc-sa-4.0
multilinguality:
ecthr:
- monolingual
scotus:
- monolingual
fscs:
- multilingual
cail:
- monolingual
size_categories:
ecthr:
- 10K<n<100K
scotus:
- 1K<n<10K
fscs:
- 10K<n<100K
cail:
- 100K<n<1M
source_datasets:
- extended
task_categories:
- text-classification
task_ids:
- multi-label-classification
- multi-class-classification
- topic-classification
pretty_name: FairLex
tags:
- bias
- gender-bias
---
# Dataset Card for "FairLex"
## 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/coastalcph/fairlex
- **Repository:** https://github.com/coastalcph/fairlex
- **Paper:** https://aclanthology.org/2022.acl-long.301/
- **Leaderboard:** -
- **Point of Contact:** [Ilias Chalkidis](mailto:ilias.chalkidis@di.ku.dk)
### Dataset Summary
We present a benchmark suite of four datasets for evaluating the fairness of pre-trained legal language models and the techniques used to fine-tune them for downstream tasks. Our benchmarks cover four jurisdictions (European Council, USA, Swiss, and Chinese), five languages (English, German, French, Italian, and Chinese), and fairness across five attributes (gender, age, nationality/region, language, and legal area). In our experiments, we evaluate pre-trained language models using several group-robust fine-tuning techniques and show that performance group disparities are vibrant in many cases, while none of these techniques guarantee fairness, nor consistently mitigate group disparities. Furthermore, we provide a quantitative and qualitative analysis of our results, highlighting open challenges in the development of robustness methods in legal NLP.
For the purpose of this work, we release four domain-specific BERT models with continued pre-training on the corpora of the examined datasets (ECtHR, SCOTUS, FSCS, CAIL). We train mini-sized BERT models with 6 Transformer blocks, 384 hidden units, and 12 attention heads. We warm-start all models from the public MiniLMv2 (Wang et al., 2021) using the distilled version of RoBERTa (Liu et al., 2019). For the English datasets (ECtHR, SCOTUS) and the one distilled from XLM-R (Conneau et al., 2021) for the rest (trilingual FSCS, and Chinese CAIL). [[Link to Models](https://huggingface.co/models?search=fairlex)]
### Supported Tasks and Leaderboards
The supported tasks are the following:
<table>
<tr><td>Dataset</td><td>Source</td><td>Sub-domain</td><td>Language</td><td>Task Type</td><td>Classes</td><tr>
<tr><td>ECtHR</td><td> <a href="https://aclanthology.org/P19-1424/">Chalkidis et al. (2019)</a> </td><td>ECHR</td><td>en</td><td>Multi-label classification</td><td>10+1</td></tr>
<tr><td>SCOTUS</td><td> <a href="http://scdb.wustl.edu">Spaeth et al. (2020)</a></td><td>US Law</td><td>en</td><td>Multi-class classification</td><td>11</td></tr>
<tr><td>FSCS</td><td> <a href="https://aclanthology.org/2021.nllp-1.3/">Niklaus et al. (2021)</a></td><td>Swiss Law</td><td>en, fr , it</td><td>Binary classification</td><td>2</td></tr>
<tr><td>CAIL</td><td> <a href="https://arxiv.org/abs/2103.13868">Wang et al. (2021)</a></td><td>Chinese Law</td><td>zh</td><td>Multi-class classification</td><td>6</td></tr>
</table>
#### ecthr
The European Court of Human Rights (ECtHR) hears allegations that a state has breached human rights provisions of the European Convention of Human Rights (ECHR). We use the dataset of Chalkidis et al. (2021), which contains 11K cases from ECtHR's public database.
Each case is mapped to *articles* of the ECHR that were violated (if any). This is a multi-label text classification task. Given the facts of a case, the goal is to predict the ECHR articles that were violated, if any, as decided (ruled) by the court. The cases are chronologically split into training (9k, 2001--16), development (1k, 2016--17), and test (1k, 2017--19) sets.
To facilitate the study of the fairness of text classifiers, we record for each case the following attributes: (a) The _defendant states_, which are the European states that allegedly violated the ECHR. The defendant states for each case is a subset of the 47 Member States of the Council of Europe; To have statistical support, we group defendant states in two groups:
Central-Eastern European states, on one hand, and all other states, as classified by the EuroVoc thesaurus. (b) The _applicant's age_ at the time of the decision. We extract the birth year of the applicant from the case facts, if possible, and classify its case in an age group (<=35, <=64, or older); and (c) the _applicant's gender_, extracted from the facts, if possible based on pronouns, classified in two categories (male, female).
#### scotus
The US Supreme Court (SCOTUS) is the highest federal court in the United States of America and generally hears only the most controversial or otherwise complex cases that have not been sufficiently well solved by lower courts.
We combine information from SCOTUS opinions with the Supreme Court DataBase (SCDB) (Spaeth, 2020). SCDB provides metadata (e.g., date of publication, decisions, issues, decision directions, and many more) for all cases. We consider the available 14 thematic issue areas (e.g, Criminal Procedure, Civil Rights, Economic Activity, etc.). This is a single-label multi-class document classification task. Given the court's opinion, the goal is to predict the issue area whose focus is on the subject matter of the controversy (dispute). SCOTUS contains a total of 9,262 cases that we split chronologically into 80% for training (7.4k, 1946--1982), 10% for development (914, 1982--1991) and 10% for testing (931, 1991--2016).
From SCDB, we also use the following attributes to study fairness: (a) the _type of respondent_, which is a manual categorization of respondents (defendants) in five categories (person, public entity, organization, facility, and other); and (c) the _direction of the decision_, i.e., whether the decision is liberal, or conservative, provided by SCDB.
#### fscs
The Federal Supreme Court of Switzerland (FSCS) is the last level of appeal in Switzerland and similarly to SCOTUS, the court generally hears only the most controversial or otherwise complex cases which have not been sufficiently well solved by lower courts. The court often focuses only on small parts of the previous decision, where they discuss possible wrong reasoning by the lower court. The Swiss-Judgment-Predict dataset (Niklaus et al., 2021) contains more than 85K decisions from the FSCS written in one of three languages (50K German, 31K French, 4K Italian) from the years 2000 to 2020.
The dataset is not parallel, i.e., all cases are unique and decisions are written only in a single language.
The dataset provides labels for a simplified binary (_approval_, _dismissal_) classification task. Given the facts of the case, the goal is to predict if the plaintiff's request is valid or partially valid. The cases are also chronologically split into training (59.7k, 2000-2014), development (8.2k, 2015-2016), and test (17.4k, 2017-2020) sets.
The dataset provides three additional attributes: (a) the _language_ of the FSCS written decision, in either German, French, or Italian; (b) the _legal area_ of the case (public, penal, social, civil, or insurance law) derived from the chambers where the decisions were heard; and (c) the _region_ that denotes in which federal region was the case originated.
#### cail
The Supreme People's Court of China (CAIL) is the last level of appeal in China and considers cases that originated from the high people's courts concerning matters of national importance. The Chinese AI and Law challenge (CAIL) dataset (Xiao et al., 2018) is a Chinese legal NLP dataset for judgment prediction and contains over 1m criminal cases. The dataset provides labels for *relevant article of criminal code* prediction, *charge* (type of crime) prediction, imprisonment *term* (period) prediction, and monetary *penalty* prediction. The publication of the original dataset has been the topic of an active debate in the NLP community(Leins et al., 2020; Tsarapatsanis and Aletras, 2021; Bender, 2021).
Recently, Wang et al. (2021) re-annotated a subset of approx. 100k cases with demographic attributes. Specifically, the new dataset has been annotated with: (a) the _applicant's gender_, classified in two categories (male, female); and (b) the _region_ of the court that denotes in which out of the 7 provincial-level administrative regions was the case judged. We re-split the dataset chronologically into training (80k, 2013-2017), development (12k, 2017-2018), and test (12k, 2018) sets. In our study, we re-frame the imprisonment _term_ prediction and examine a soft version, dubbed _crime severity_ prediction task, a multi-class classification task, where given the facts of a case, the goal is to predict how severe was the committed crime with respect to the imprisonment term. We approximate crime severity by the length of imprisonment term, split in 6 clusters (0, <=12, <=36, <=60, <=120, >120 months).
### Languages
We consider datasets in English, German, French, Italian, and Chinese.
## Dataset Structure
### Data Instances
#### ecthr
An example of 'train' looks as follows.
```json
{
"text": "1. At the beginning of the events relevant to the application, K. had a daughter, P., and a son, M., born in 1986 and 1988 respectively. ... ",
"labels": [4],
"defendant_state": 1,
"applicant_gender": 0,
"applicant_age": 0
}
```
#### scotus
An example of 'train' looks as follows.
```json
{
"text": "United States Supreme Court MICHIGAN NAT. BANK v. MICHIGAN(1961) No. 155 Argued: Decided: March 6, 1961 </s> R. S. 5219 permits States to tax the shares of national banks, but not at a greater rate than . . . other moneyed capital . . . coming into competition with the business of national banks ...",
"label": 9,
"decision_direction": 0,
"respondent_type": 3
}
```
#### fscs
An example of 'train' looks as follows.
```json
{
"text": "A.- Der 1955 geborene V._ war seit 1. September 1986 hauptberuflich als technischer Kaufmann bei der Firma A._ AG tätig und im Rahmen einer Nebenbeschäftigung (Nachtarbeit) ab Mai 1990 bei einem Bewachungsdienst angestellt gewesen, als er am 10....",
"label": 0,
"decision_language": 0,
"legal_are": 5,
"court_region": 2
}
```
#### cail
An example of 'train' looks as follows.
```json
{
"text": "南宁市兴宁区人民检察院指控，2012年1月1日19时许，被告人蒋满德在南宁市某某路某号某市场内，因经营问题与被害人杨某某发生争吵并推打 ...",
"label": 0,
"defendant_gender": 0,
"court_region": 5
}
```
### Data Fields
#### ecthr_a
- `text`: a `string` feature (factual paragraphs (facts) from the case description).
- `labels`: a list of classification labels (a list of violated ECHR articles, if any). The ECHR articles considered are 2, 3, 5, 6, 8, 9, 11, 14, P1-1.
- `defendant_state`: Defendant State group (C.E. European, Rest of Europe)
- `applicant_gender`: The gender of the applicant (N/A, Male, Female)
- `applicant_age`: The age group of the applicant (N/A, <=35, <=64, or older)
#### scotus
- `text`: a `string` feature (the court opinion).
- `label`: a classification label (the relevant issue area). The issue areas are: (1, Criminal Procedure), (2, Civil Rights), (3, First Amendment), (4, Due Process), (5, Privacy), (6, Attorneys), (7, Unions), (8, Economic Activity), (9, Judicial Power), (10, Federalism), (11, Interstate Relations), (12, Federal Taxation), (13, Miscellaneous), (14, Private Action).
- `respondent_type`: the type of respondent, which is a manual categorization (clustering) of respondents (defendants) in five categories (person, public entity, organization, facility, and other).
- `decision_direction`: the direction of the decision, i.e., whether the decision is liberal, or conservative, provided by SCDB.
#### fscs
- `text`: a `string` feature (an EU law).
- `label`: a classification label (approval or dismissal of the appeal).
- `language`: the language of the FSCS written decision, (German, French, or Italian).
- `legal_area`: the legal area of the case (public, penal, social, civil, or insurance law) derived from the chambers where the decisions were heard.
- `region`: the region that denotes in which federal region was the case originated.
#### cail
- `text`: a `string` feature (the factual description of the case).
- `label`: a classification label (crime severity derived by the imprisonment term).
- `defendant_gender`: the gender of the defendant (Male or Female).
- `court_region`: the region of the court that denotes in which out of the 7 provincial-level administrative regions was the case judged.
### Data Splits
<table>
<tr><td>Dataset </td><td>Training</td><td>Development</td><td>Test</td><td>Total</td></tr>
<tr><td>ECtHR</td><td>9000</td><td>1000</td><td>1000</td><td>11000</td></tr>
<tr><td>SCOTUS</td><td>7417</td><td>914</td><td>931</td><td>9262</td></tr>
<tr><td>FSCS</td><td>59709</td><td>8208</td><td>17357</td><td>85274</td></tr>
<tr><td>CAIL</td><td>80000</td><td>12000</td><td>12000</td><td>104000</td></tr>
</table>
## 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
<table>
<tr><td>Dataset</td><td>Source</td><td>Sub-domain</td><td>Language</td><td>Task Type</td><td>Classes</td><tr>
<tr><td>ECtHR</td><td> <a href="https://aclanthology.org/P19-1424/">Chalkidis et al. (2019)</a> </td><td>ECHR</td><td>en</td><td>Multi-label classification</td><td>10+1</td></tr>
<tr><td>SCOTUS</td><td> <a href="http://scdb.wustl.edu">Spaeth et al. (2020)</a></td><td>US Law</td><td>en</td><td>Multi-class classification</td><td>14</td></tr>
<tr><td>FSCS</td><td> <a href="https://aclanthology.org/2021.nllp-1.3/">Niklaus et al. (2021)</a></td><td>Swiss Law</td><td>en, fr , it</td><td>Binary classification</td><td>2</td></tr>
<tr><td>CAIL</td><td> <a href="https://arxiv.org/abs/2105.03887">Wang et al. (2021)</a></td><td>Chinese Law</td><td>zh</td><td>Multi-class classification</td><td>6</td></tr>
</table>
#### Initial Data Collection and Normalization
We standardize and put together four datasets: ECtHR (Chalkidis et al., 2021), SCOTUS (Spaeth et al., 2020), FSCS (Niklaus et al., 2021), and CAIL (Xiao et al., 2018; Wang et al., 2021) that are already publicly available.
The benchmark is not a blind stapling of pre-existing resources, we augment previous datasets. In the case of ECtHR, previously unavailable demographic attributes have been released to make the original dataset amenable for fairness research. For SCOTUS, two resources (court opinions with SCDB) have been combined for the very same reason, while the authors provide a manual categorization (clustering) of respondents.
All datasets, except SCOTUS, are publicly available and have been previously published. If datasets or the papers where they were introduced were not compiled or written by the authors, the original work is referenced and authors encourage FairLex users to do so as well. In fact, this work should only be referenced, in addition to citing the original work, when jointly experimenting with multiple FairLex datasets and using the FairLex evaluation framework and infrastructure, or using any newly introduced annotations (ECtHR, SCOTUS). Otherwise only the original work should be cited.
#### 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?
All classification labels rely on legal decisions (ECtHR, FSCS, CAIL), or are part of archival procedures (SCOTUS).
The demographic attributes and other metadata are either provided by the legal databases or have been extracted automatically from the text by means of Regular Expressions.
Consider the **Dataset Description** and **Discussion of Biases** sections, and the original publication for detailed information.
### Personal and Sensitive Information
The data is in general partially anonymized in accordance with the applicable national law. The data is considered to be in the public sphere from a privacy perspective. This is a very sensitive matter, as the courts try to keep a balance between transparency (the public's right to know) and privacy (respect for private and family life).
ECtHR cases are partially annonymized by the court. Its data is processed and made public in accordance with the European Data Protection Law.
SCOTUS cases may also contain personal information and the data is processed and made available by the US Supreme Court, whose proceedings are public. While this ensures compliance with US law, it is very likely that similarly to the ECtHR any processing could be justified by either implied consent or legitimate interest under European law. In FSCS, the names of the parties have been redacted by the courts according to the official guidelines. CAIL cases are also partially anonymized by the courts according to the courts' policy. Its data is processed and made public in accordance with Chinese Law.
## Considerations for Using the Data
### Social Impact of Dataset
This work can help practitioners to build assisting technology for legal professionals - with respect to the legal framework (jurisdiction) they operate -; technology that does not only rely on performance on majority groups but also considering minorities and the robustness of the developed models across them. This is an important application field, where more research should be conducted (Tsarapatsanis and Aletras, 2021) in order to improve legal services and democratize law, but more importantly, highlight (inform the audience on) the various multi-aspect shortcomings seeking a responsible and ethical (fair) deployment of technology.
### Discussion of Biases
The current version of FairLex covers a very small fraction of legal applications, jurisdictions, and protected attributes. The benchmark inevitably cannot cover "_everything in the whole wide (legal) world_" (Raji et al., 2021), but nonetheless, we believe that the published resources will help critical research in the area of fairness.
Some protected attributes within the datasets are extracted automatically, i.e., the gender and the age of the ECtHR dataset, by means of Regular Expressions, or manually clustered by the authors, such as the defendant state in the ECtHR dataset and the respondent attribute in the SCOTUS dataset. Those assumptions and simplifications can hold in an experimental setting only and by no means should be used in real-world applications where some simplifications, e.g., binary gender, would not be appropriate. By no means, do the authors or future users have to endorse the law standards or framework of the examined datasets, to any degree rather than the publication and use of the data.
### Other Known Limitations
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
## Additional Information
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
### Dataset Curators
*Ilias Chalkidis, Tommaso Pasini, Sheng Zhang, Letizia Tomada, Letizia, Sebastian Felix Schwemer, Anders Søgaard.*
*FairLex: A Multilingual Benchmark for Evaluating Fairness in Legal Text Processing.*
*2022. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics, Dublin, Ireland.*
**Note:** The original datasets have been originally curated by others, and further curated (updated) by means of this benchmark.
### Licensing Information
The benchmark is released under a [Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0)](https://creativecommons.org/licenses/by-nc-sa/4.0/) license. The licensing is compatible with the licensing of former material (remixed, transformed datasets).
### Citation Information
[*Ilias Chalkidis, Tommaso Pasini, Sheng Zhang, Letizia Tomada, Letizia, Sebastian Felix Schwemer, Anders Søgaard.*
*FairLex: A Multilingual Benchmark for Evaluating Fairness in Legal Text Processing.*
*2022. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics, Dublin, Ireland.*](https://aclanthology.org/2022.acl-long.301/)
```
@inproceedings{chalkidis-etal-2022-fairlex,
author={Chalkidis, Ilias and Passini, Tommaso and Zhang, Sheng and
Tomada, Letizia and Schwemer, Sebastian Felix and Søgaard, Anders},
title={FairLex: A Multilingual Benchmark for Evaluating Fairness in Legal Text Processing},
booktitle={Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics},
year={2022},
address={Dublin, Ireland}
}
```
**Note:** Please consider citing and giving credits to all publications releasing the examined datasets.
### Contributions
Thanks to [@iliaschalkidis](https://github.com/iliaschalkidis) for adding this dataset.
|
codeceejay | null | null | null | false | 165 | false | codeceejay/ng_accent | 2022-01-28T16:41:32.000Z | null | false | 6e4bef0cfa6a9570ba29b06ca47a2db111f71cc0 | [] | [] | https://huggingface.co/datasets/codeceejay/ng_accent/resolve/main/README.md | |
cointegrated | null | null | null | false | 322 | false | cointegrated/ru-paraphrase-NMT-Leipzig | 2022-10-23T12:23:15.000Z | null | false | 9070da7298a73ea6129f711916f17e52d82884de | [] | [
"annotations_creators:no-annotation",
"language_creators:machine-generated",
"language:ru",
"license:cc-by-4.0",
"multilinguality:translation",
"size_categories:100K<n<1M",
"source_datasets:extended|other",
"task_categories:text-generation",
"tags:conditional-text-generation",
"tags:paraphrase-generation",
"tags:paraphrase"
] | https://huggingface.co/datasets/cointegrated/ru-paraphrase-NMT-Leipzig/resolve/main/README.md | ---
annotations_creators:
- no-annotation
language_creators:
- machine-generated
language:
- ru
license:
- cc-by-4.0
multilinguality:
- translation
size_categories:
- 100K<n<1M
source_datasets:
- extended|other
task_categories:
- text-generation
pretty_name: ru-paraphrase-NMT-Leipzig
tags:
- conditional-text-generation
- paraphrase-generation
- paraphrase
---
# Dataset Card for **cointegrated/ru-paraphrase-NMT-Leipzig**
## 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
- **Paper:** https://habr.com/ru/post/564916/
- **Point of Contact:** [@cointegrated](https://huggingface.co/cointegrated)
### Dataset Summary
The dataset contains 1 million Russian sentences and their automatically generated paraphrases.
It was created by David Dale ([@cointegrated](https://huggingface.co/cointegrated)) by translating the `rus-ru_web-public_2019_1M` corpus from [the Leipzig collection](https://wortschatz.uni-leipzig.de/en/download) into English and back into Russian. A fraction of the resulting paraphrases are invalid, and should be filtered out.
The blogpost ["Перефразирование русских текстов: корпуса, модели, метрики"](https://habr.com/ru/post/564916/) provides a detailed description of the dataset and its properties.
The dataset can be loaded with the following code:
```Python
import datasets
data = datasets.load_dataset(
'cointegrated/ru-paraphrase-NMT-Leipzig',
data_files={"train": "train.csv","val": "val.csv","test": "test.csv"},
)
```
Its output should look like
```
DatasetDict({
train: Dataset({
features: ['idx', 'original', 'en', 'ru', 'chrf_sim', 'labse_sim'],
num_rows: 980000
})
val: Dataset({
features: ['idx', 'original', 'en', 'ru', 'chrf_sim', 'labse_sim'],
num_rows: 10000
})
test: Dataset({
features: ['idx', 'original', 'en', 'ru', 'chrf_sim', 'labse_sim'],
num_rows: 10000
})
})
```
### Supported Tasks and Leaderboards
The dataset can be used to train and validate models for paraphrase generation or (if negative sampling is used) for paraphrase detection.
### Languages
Russian (main), English (auxilliary).
## Dataset Structure
### Data Instances
Data instances look like
```
{
"labse_sim": 0.93502015,
"chrf_sim": 0.4946451012684782,
"idx": 646422,
"ru": "О перспективах развития новых медиа-технологий в РФ расскажут на медиафоруме Енисея.",
"original": "Перспективы развития новых медиатехнологий в Российской Федерации обсудят участники медиафорума «Енисей.",
"en": "Prospects for the development of new media technologies in the Russian Federation will be discussed at the Yenisey Media Forum."
}
```
Where `original` is the original sentence, and `ru` is its machine-generated paraphrase.
### Data Fields
- `idx`: id of the instance in the original corpus
- `original`: the original sentence
- `en`: automatic translation of `original` to English
- `ru`: automatic translation of `en` back to Russian, i.e. a paraphrase of `original`
- `chrf_sim`: [ChrF++](https://huggingface.co/metrics/chrf) similarity of `original` and `ru`
- `labse_sim`: cosine similarity of [LaBSE](https://huggingface.co/cointegrated/LaBSE-en-ru) embedings of `original` and `ru`
- `forward_entailment`: predicted probability that `original` entails `ru`
- `backward_entailment`: predicted probability that `ru` entails `original`
- `p_good`: predicted probability that `ru` and `original` have equivalent meaning
### Data Splits
Train – 980K, validation – 10K, test – 10K. The splits were generated randomly.
## Dataset Creation
### Curation Rationale
There are other Russian paraphrase corpora, but they have major drawbacks:
- The best known [corpus from paraphraser.ru 2016 contest](http://paraphraser.ru/download/) is rather small and covers only the News domain.
- [Opusparcus](https://huggingface.co/datasets/GEM/opusparcus), [ParaPhraserPlus](http://paraphraser.ru/download/), and [corpora of Tamara Zhordanija](https://github.com/tamriq/paraphrase) are noisy, i.e. a large proportion of sentence pairs in them have substantial difference in meaning.
- The Russian part of [TaPaCo](https://huggingface.co/datasets/tapaco) has very high lexical overlap in the sentence pairs; in other words, their paraphrases are not diverse enough.
The current corpus is generated with a dual objective: the parphrases should be semantically as close as possible to the original sentences, while being lexically different from them. Back-translation with restricted vocabulary seems to achieve this goal often enough.
### Source Data
#### Initial Data Collection and Normalization
The `rus-ru_web-public_2019_1M` corpus from [the Leipzig collection](https://wortschatz.uni-leipzig.de/en/download) as is.
The process of its creation is described [in this paper](http://www.lrec-conf.org/proceedings/lrec2012/pdf/327_Paper.pdf):
D. Goldhahn, T. Eckart & U. Quasthoff: Building Large Monolingual Dictionaries at the Leipzig Corpora Collection: From 100 to 200 Languages.
In: *Proceedings of the 8th International Language Resources and Evaluation (LREC'12), 2012*.
#### Automatic paraphrasing
The paraphrasing was carried out by translating the original sentence to English and then back to Russian.
The models [facebook/wmt19-ru-en](https://huggingface.co/facebook/wmt19-ru-en) and [facebook/wmt19-en-ru](https://huggingface.co/facebook/wmt19-en-ru) were used for translation.
To ensure that the back-translated texts are not identical to the original texts, the final decoder was prohibited to use the token n-grams from the original texts.
The code below implements the paraphrasing function.
```python
import torch
from transformers import FSMTModel, FSMTTokenizer, FSMTForConditionalGeneration
tokenizer = FSMTTokenizer.from_pretrained("facebook/wmt19-en-ru")
model = FSMTForConditionalGeneration.from_pretrained("facebook/wmt19-en-ru")
inverse_tokenizer = FSMTTokenizer.from_pretrained("facebook/wmt19-ru-en")
inverse_model = FSMTForConditionalGeneration.from_pretrained("facebook/wmt19-ru-en")
model.cuda();
inverse_model.cuda();
def paraphrase(text, gram=4, num_beams=5, **kwargs):
""" Generate a paraphrase using back translation.
Parameter `gram` denotes size of token n-grams of the original sentence that cannot appear in the paraphrase.
"""
input_ids = inverse_tokenizer.encode(text, return_tensors="pt")
with torch.no_grad():
outputs = inverse_model.generate(input_ids.to(inverse_model.device), num_beams=num_beams, **kwargs)
other_lang = inverse_tokenizer.decode(outputs[0], skip_special_tokens=True)
# print(other_lang)
input_ids = input_ids[0, :-1].tolist()
bad_word_ids = [input_ids[i:(i+gram)] for i in range(len(input_ids)-gram)]
input_ids = tokenizer.encode(other_lang, return_tensors="pt")
with torch.no_grad():
outputs = model.generate(input_ids.to(model.device), num_beams=num_beams, bad_words_ids=bad_word_ids, **kwargs)
decoded = tokenizer.decode(outputs[0], skip_special_tokens=True)
return decoded
```
The corpus was created by running the above `paraphrase` function on the original sentences with parameters `gram=3, num_beams=5, repetition_penalty=3.14, no_repeat_ngram_size=6`.
### Annotations
#### Annotation process
The dataset was annotated by several automatic metrics:
- [ChrF++](https://huggingface.co/metrics/chrf) between `original` and `ru` sentences;
- cosine similarity between [LaBSE](https://huggingface.co/cointegrated/LaBSE-en-ru) embeddings of these sentences;
- forward and backward entailment probabilites predictd by the [rubert-base-cased-nli-twoway](https://huggingface.co/cointegrated/rubert-base-cased-nli-twoway) model;
- `p_good`, a metric aggregating the four metrics above into a single number. It is obtained with a logistic regression trained on 100 randomly chosen from the train set and manually labelled sentence pairs.
#### Who are the annotators?
Human annotation was involved only for a small subset used to train the model for `p_good`. It was conduced by the dataset author, @cointegrated.
### Personal and Sensitive Information
The dataset is not known to contain any personal or sensitive information.
The sources and processes of original data collection are described at https://wortschatz.uni-leipzig.de/en/download.
## Considerations for Using the Data
### Social Impact of Dataset
The dataset may enable creation for paraphrasing systems that can be used both for "good" purposes (such as assisting writers or augmenting text datasets), and for "bad" purposes (such as disguising plagiarism). The authors are not responsible for any uses of the dataset.
### Discussion of Biases
The dataset may inherit some of the biases of [the underlying Leipzig web corpus](https://wortschatz.uni-leipzig.de/en/download) or the neural machine translation models ([1](https://huggingface.co/facebook/wmt19-ru-en), [2](https://huggingface.co/facebook/wmt19-en-ru)) with which it was generated.
### Other Known Limitations
Most of the paraphrases in the dataset are valid (by a rough estimante, at least 80%). However, in some sentence pairs there are faults:
- Named entities are often spelled in different ways (e.g. `"Джейкоб" -> "Яков") or even replaced with other entities (e.g. `"Оймякон" -> "Оймянск" or `"Верхоянск" -> "Тольятти"`).
- Sometimes the meaning of words or phrases changes signigicantly, e.g. `"полустанок" -> "полумашина"`, or `"были по колено в грязи" -> "лежали на коленях в иле"`.
- Sometimes the syntax is changed in a meaning-altering way, e.g. `"Интеллектуальное преимущество Вавилова и его соратников над демагогами из рядов сторонников новой агробиологии разительно очевидно." -> "Интеллектуал Вавилов и его приспешники в новой аграрной биологии явно превзошли демогогов."`.
- Grammatical properties that are present in Russian morphology but absent in English, such as gender, are often lost, e.g. `"Я не хотела тебя пугать" -> "Я не хотел пугать вас"`.
The field `labse_sim` reflects semantic similarity between the sentences, and it can be used to filter out at least some poor paraphrases.
## Additional Information
### Dataset Curators
The dataset was created by [David Dale](https://daviddale.ru/en), a.k.a. [@cointegrated](https://huggingface.co/cointegrated).
### Licensing Information
This corpus, as well as the original Leipzig corpora, are licensed under [CC BY](http://creativecommons.org/licenses/by/4.0/).
### Citation Information
[This blog post](https://habr.com/ru/post/564916/) can be cited:
```
@misc{dale_paraphrasing_2021,
author = "Dale, David",
title = "Перефразирование русских текстов: корпуса, модели, метрики",
editor = "habr.com",
url = "https://habr.com/ru/post/564916/",
month = {June},
year = {2021},
note = {[Online; posted 28-June-2021]},
}
```
### Contributions
Thanks to [@avidale](https://github.com/avidale) for adding this dataset. |
collectivat | null | @inproceedings{kulebi18_iberspeech,
author={Baybars Külebi and Alp Öktem},
title={{Building an Open Source Automatic Speech Recognition System for Catalan}},
year=2018,
booktitle={Proc. IberSPEECH 2018},
pages={25--29},
doi={10.21437/IberSPEECH.2018-6}
} | This corpus includes 240 hours of Catalan speech from broadcast material.
The details of segmentation, data processing and also model training are explained in Külebi, Öktem; 2018.
The content is owned by Corporació Catalana de Mitjans Audiovisuals, SA (CCMA);
we processed their material and hereby making it available under their terms of use.
This project was supported by the Softcatalà Association. | false | 320 | false | collectivat/tv3_parla | 2022-10-25T11:46:40.000Z | null | false | c7c41c1de61a15c5e4990b3574c2c3baa2119e41 | [] | [
"annotations_creators:found",
"language_creators:found",
"language:ca",
"license:cc-by-nc-4.0",
"multilinguality:monolingual",
"size_categories:100K<n<1M",
"source_datasets:original",
"task_categories:automatic-speech-recognition",
"task_categories:text-generation",
"task_ids:language-modeling"
] | https://huggingface.co/datasets/collectivat/tv3_parla/resolve/main/README.md | ---
annotations_creators:
- found
language_creators:
- found
language:
- ca
license:
- cc-by-nc-4.0
multilinguality:
- monolingual
size_categories:
- 100K<n<1M
source_datasets:
- original
task_categories:
- automatic-speech-recognition
- text-generation
task_ids:
- language-modeling
pretty_name: TV3Parla
---
# Dataset Card for TV3Parla
## 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:** https://collectivat.cat/asr#tv3parla
- **Repository:**
- **Paper:** [Building an Open Source Automatic Speech Recognition System for Catalan](https://www.isca-speech.org/archive/iberspeech_2018/kulebi18_iberspeech.html)
- **Point of Contact:** [Col·lectivaT](mailto:info@collectivat.cat)
### Dataset Summary
This corpus includes 240 hours of Catalan speech from broadcast material.
The details of segmentation, data processing and also model training are explained in Külebi, Öktem; 2018.
The content is owned by Corporació Catalana de Mitjans Audiovisuals, SA (CCMA);
we processed their material and hereby making it available under their terms of use.
This project was supported by the Softcatalà Association.
### Supported Tasks and Leaderboards
The dataset can be used for:
- Language Modeling.
- Automatic Speech Recognition (ASR) transcribes utterances into words.
### Languages
The dataset is in Catalan (`ca`).
## Dataset Structure
### Data Instances
```
{
'path': 'tv3_0.3/wav/train/5662515_1492531876710/5662515_1492531876710_120.180_139.020.wav',
'audio': {'path': 'tv3_0.3/wav/train/5662515_1492531876710/5662515_1492531876710_120.180_139.020.wav',
'array': array([-0.01168823, 0.01229858, 0.02819824, ..., 0.015625 ,
0.01525879, 0.0145874 ]),
'sampling_rate': 16000},
'text': 'algunes montoneres que que et feien anar ben col·locat i el vent també hi jugava una mica de paper bufava vent de cantó alguns cops o de cul i el pelotón el vent el porta molt malament hi havia molts nervis'
}
```
### Data Fields
- `path` (str): Path to the audio file.
- `audio` (dict): A dictionary containing the path to the downloaded audio file, the decoded audio array, and the sampling
rate. Note that when accessing the audio column: `dataset[0]["audio"]` the audio file is automatically decoded and
resampled to `dataset.features["audio"].sampling_rate`. Decoding and resampling of a large number of audio files might
take a significant amount of time. Thus, it is important to first query the sample index before the `"audio"` column,
*i.e.* `dataset[0]["audio"]` should **always** be preferred over `dataset["audio"][0]`.
- `text` (str): Transcription of the audio file.
### Data Splits
The dataset is split into "train" and "test".
| | train | test |
|:-------------------|-------:|-----:|
| Number of examples | 159242 | 2220 |
## 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
[Creative Commons Attribution-NonCommercial 4.0 International](https://creativecommons.org/licenses/by-nc/4.0/).
### Citation Information
```
@inproceedings{kulebi18_iberspeech,
author={Baybars Külebi and Alp Öktem},
title={{Building an Open Source Automatic Speech Recognition System for Catalan}},
year=2018,
booktitle={Proc. IberSPEECH 2018},
pages={25--29},
doi={10.21437/IberSPEECH.2018-6}
}
```
### Contributions
Thanks to [@albertvillanova](https://github.com/albertvillanova) for adding this dataset.
|
comodoro | null | null | null | false | 319 | false | comodoro/pscr | 2022-02-08T07:07:49.000Z | null | false | 853146eccb23be28175456f81456e82cba2f83f1 | [] | [
"license:cc-by-nc-3.0"
] | https://huggingface.co/datasets/comodoro/pscr/resolve/main/README.md | ---
license: cc-by-nc-3.0
---
|
comodoro | null | @misc{11234/1-1740,
title = {Vystadial 2016 – Czech data},
author = {Pl{\'a}tek, Ond{\v r}ej and Du{\v s}ek, Ond{\v r}ej and Jur{\v c}{\'{\i}}{\v c}ek, Filip},
url = {http://hdl.handle.net/11234/1-1740},
note = {{LINDAT}/{CLARIAH}-{CZ} digital library at the Institute of Formal and Applied Linguistics ({{\'U}FAL}), Faculty of Mathematics and Physics, Charles University},
copyright = {Creative Commons - Attribution-{ShareAlike} 4.0 International ({CC} {BY}-{SA} 4.0)},
year = {2016} } | This is the Czech data collected during the `VYSTADIAL` project. It is an extension of the 'Vystadial 2013' Czech part data release. The dataset comprises of telephone conversations in Czech, developed for training acoustic models for automatic speech recognition in spoken dialogue systems. | false | 319 | false | comodoro/vystadial2016_asr | 2022-09-02T08:41:16.000Z | null | false | 219094aed954b897758697a8921a854f5e199b70 | [] | [
"license:cc-by-nc-3.0"
] | https://huggingface.co/datasets/comodoro/vystadial2016_asr/resolve/main/README.md | ---
license: cc-by-nc-3.0
---
|
corypaik | null | @misc{paik2021world,
title={The World of an Octopus: How Reporting Bias Influences a Language Model's Perception of Color},
author={Cory Paik and Stéphane Aroca-Ouellette and Alessandro Roncone and Katharina Kann},
year={2021},
eprint={2110.08182},
archivePrefix={arXiv},
primaryClass={cs.CL}
} | *The Color Dataset* (CoDa) is a probing dataset to evaluate the representation of visual properties in language models. CoDa consists of color distributions for 521 common objects, which are split into 3 groups: Single, Multi, and Any. | false | 518 | false | corypaik/coda | 2022-10-20T16:57:23.000Z | coda | false | 9f47e7ea19a1f969027a138c92e4e3a71b5537d3 | [] | [
"arxiv:2110.08182",
"annotations_creators:crowdsourced",
"language_creators:expert-generated",
"language:en",
"language_bcp47:en-US",
"license:apache-2.0",
"multilinguality:monolingual",
"size_categories:10K<n<100K",
"source_datasets:original",
"task_ids:text-scoring-other-distribution-prediction"
] | https://huggingface.co/datasets/corypaik/coda/resolve/main/README.md | ---
annotations_creators:
- crowdsourced
language_creators:
- expert-generated
language:
- en
language_bcp47:
- en-US
license:
- apache-2.0
multilinguality:
- monolingual
pretty_name: CoDa
paperswithcode_id: coda
size_categories:
- 10K<n<100K
source_datasets:
- original
task_categories:
- text-scoring
task_ids:
- text-scoring-other-distribution-prediction
---
# Dataset Card for CoDa
## 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
- **Repository:** [nala-cub/coda](https://github.com/nala-cub/coda)
- **Paper:** [The World of an Octopus: How Reporting Bias Influences a Language Model's Perception of Color](https://arxiv.org/abs/2110.08182)
- **Point of Contact:** [Cory Paik](cory.paik@colorado.edu)
### Dataset Summary
*The Color Dataset* (CoDa) is a probing dataset to evaluate the representation of visual properties in language models. CoDa consists of color distributions for 521 common objects, which are split into 3 groups. We denote these groups as Single, Multi, and Any, which represents the typical object of each group.
The default configuration of CoDa uses 10 CLIP-style templates (e.g. "A photo of a [object]"), and 10 cloze-style templates (e.g. "Everyone knows most [object] are
[color]." )
### Supported Tasks and Leaderboards
This version of the dataset consists of the filtered and templated examples as cloze style questions. See the [GitHub](https://github.com/nala-cub/coda) repo for the raw data (e.g. unfiltered annotations) as well as example usage with GPT-2, RoBERTa, ALBERT, and CLIP.
### Languages
The text in the dataset is in English. The associated BCP-47 code is `en-US`.
## Dataset Structure
### Data Instances
An example looks like this:
```json
{
"text": "All rulers are [MASK].",
"label": [
0.0181818176, 0.0363636352, 0.3077272773, 0.0181818176, 0.0363636352,
0.086363636, 0.0363636352, 0.0363636352, 0.0363636352, 0.086363636,
0.301363647
],
"template_group": 1,
"template_idx": 0,
"class_id": "/m/0hdln",
"display_name": "Ruler",
"object_group": 2,
"ngram": "ruler"
}
```
### Data Fields
- `text`: The templated example. What this is depends on the value of `template_group`.
- `template_group=0`: A CLIP style example. There are no `[MASK]` tokens in these examples.
- `template_group=1`: A cloze style example. Note that all templates have `[MASK]` as the last word, but in most cases, the period should be included.
- `label`: A list of probability values for the 11 colors. Note that these are sorted by the alphabetic order of the 11 colors (black, blue, brown, gray, green, orange, pink, purple, red, white, yellow).
- `template_group`: Type of template, `0` corresponds to A CLIP style template (`clip-imagenet`), and `1` corresponds to A cloze style templates (`text-masked`).
- `template_idx`: The index of the template out of all templates
- `class_id`: The Corresponding [OpenImages v6](https://storage.googleapis.com/openimages/web/index.html) `ClassID`.
- `display_name`: The Corresponding [OpenImages v6](https://storage.googleapis.com/openimages/web/index.html) `DisplayName`.
- `object_group`: Object Group, values correspond to `Single`, `Multi`, and `Any`.
- `ngram`: Corresponding n-gram used for lookups.
### Data Splits
Object Splits:
| Group | All | Train | Valid | Test |
| ------ | --- | ----- | ----- | ---- |
| Single | 198 | 118 | 39 | 41 |
| Multi | 208 | 124 | 41 | 43 |
| Any | 115 | 69 | 23 | 23 |
| Total | 521 | 311 | 103 | 107 |
Example Splits:
| Group | All | Train | Valid | Test |
| ------ | ----- | ----- | ----- | ---- |
| Single | 3946 | 2346 | 780 | 820 |
| Multi | 4146 | 2466 | 820 | 860 |
| Any | 2265 | 1352 | 460 | 453 |
| Total | 10357 | 6164 | 2060 | 2133 |
## 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
CoDa is licensed under the Apache 2.0 license.
### Citation Information
```
@misc{paik2021world,
title={The World of an Octopus: How Reporting Bias Influences a Language Model's Perception of Color},
author={Cory Paik and Stéphane Aroca-Ouellette and Alessandro Roncone and Katharina Kann},
year={2021},
eprint={2110.08182},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
```
### Contributions
Thanks to [@github-username](https://github.com/<github-username>) for adding this dataset.
|
corypaik | null | @inproceedings{aroca-ouellette-etal-2021-prost,
title = "{PROST}: {P}hysical Reasoning about Objects through Space and Time",
author = "Aroca-Ouellette, St{\'e}phane and
Paik, Cory and
Roncone, Alessandro and
Kann, Katharina",
booktitle = "Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021",
month = aug,
year = "2021",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.findings-acl.404",
pages = "4597--4608",
} | *Physical Reasoning about Objects Through Space and Time* (PROST) is a probing dataset to evaluate the ability of pretrained LMs to understand and reason about the physical world. PROST consists of 18,736 cloze-style multiple choice questions from 14 manually curated templates, covering 10 physical reasoning concepts: direction, mass, height, circumference, stackable, rollable, graspable, breakable, slideable, and bounceable. | false | 779 | false | corypaik/prost | 2022-10-25T09:07:34.000Z | prost | false | b3efebf08969fc19335ba894353316878b6fa493 | [] | [
"arxiv:2106.03634",
"annotations_creators:expert-generated",
"extended:original",
"language_creators:expert-generated",
"language:en-US",
"license:apache-2.0",
"multilinguality:monolingual",
"size_categories:10K<n<100K",
"source_datasets:original",
"task_categories:question-answering",
"task_ids:multiple-choice-qa",
"task_ids:open-domain-qa"
] | https://huggingface.co/datasets/corypaik/prost/resolve/main/README.md | ---
annotations_creators:
- expert-generated
extended:
- original
language_creators:
- expert-generated
language:
- en-US
license:
- apache-2.0
multilinguality:
- monolingual
paperswithcode_id: prost
size_categories:
- 10K<n<100K
source_datasets:
- original
task_categories:
- question-answering
task_ids:
- multiple-choice-qa
- open-domain-qa
---
# PROST: Physical Reasoning about Objects Through Space and Time
## 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/nala-cub/prost
- **Paper:** https://arxiv.org/abs/2106.03634
- **Leaderboard:**
- **Point of Contact:** [Stéphane Aroca-Ouellette](mailto:stephane.aroca-ouellette@colorado.edu)
### Dataset Summary
*Physical Reasoning about Objects Through Space and Time* (PROST) is a probing dataset to evaluate the ability of pretrained LMs to understand and reason about the physical world. PROST consists of 18,736 cloze-style multiple choice questions from 14 manually curated templates, covering 10 physical reasoning concepts: direction, mass, height, circumference, stackable, rollable, graspable, breakable, slideable, and bounceable.
### Supported Tasks and Leaderboards
The task is multiple choice question answering, but you can formulate it multiple ways. You can use `context` and `question` to form cloze style questions, or `context` and `ex_question` as multiple choice question answering. See the [GitHub](https://github.com/nala-cub/prost) repo for examples using GPT-1, GPT-2, BERT, RoBERTa, ALBERT, T5, and UnifiedQA.
### Languages
The text in the dataset is in English. The associated BCP-47 code is `en-US`.
## Dataset Structure
### Data Instances
An example looks like this:
```json
{
"A": "glass",
"B": "pillow",
"C": "coin",
"D": "ball",
"context": "A person drops a glass, a pillow, a coin, and a ball from a balcony.",
"ex_question": "Which object is the most likely to break?",
"group": "breaking",
"label": 0,
"name": "breaking_1",
"question": "The [MASK] is the most likely to break."
}
```
### Data Fields
- `A`: Option A (0)
- `B`: Option B (1)
- `C`: Option C (2)
- `D`: Option D (3)
- `context`: Context for the question
- `question`: A cloze style continuation of the context.
- `ex_question`: A multiple-choice style question.
- `group`: The question group, e.g. *bouncing*
- `label`: A ClassLabel indication the correct option
- `name':` The template identifier.
### Data Splits
The dataset contains 18,736 examples for testing.
## Dataset Creation
### Curation Rationale
PROST is designed to avoid models succeeding in unintended ways. First, PROST provides no training data, so as to probe models in a zero-shot fashion. This prevents models from succeeding through spurious correlations between testing and training, and encourages success through a true understanding of and reasoning about the concepts at hand. Second, we manually write templates for all questions in an effort to prevent models from having seen the exact same sentences in their training data. Finally, it focuses on a small set of well defined, objective concepts that only require a small vocabulary. This allows researchers to focus more on the quality of training data rather than on size of it.
### 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
PROST is licensed under the Apache 2.0 license.
### Citation Information
```
@inproceedings{aroca-ouellette-etal-2021-prost,
title = "{PROST}: {P}hysical Reasoning about Objects through Space and Time",
author = "Aroca-Ouellette, St{\'e}phane and
Paik, Cory and
Roncone, Alessandro and
Kann, Katharina",
booktitle = "Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021",
month = aug,
year = "2021",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.findings-acl.404",
pages = "4597--4608",
}
```
### Contributions
Thanks to [@corypaik](https://github.com/corypaik) for adding this dataset.
|
coyotte508 | null | null | null | false | 322 | false | coyotte508/dataset | 2022-07-08T11:20:09.000Z | wider-face-1 | false | 9dd3cd583a7f1f4400a53ab0c9fc4b3dec4d5071 | [] | [
"arxiv:1511.06523",
"annotations_creators:expert-generated",
"language_creators:found",
"language:en",
"license:cc-by-nc-nd-4.0",
"multilinguality:monolingual",
"size_categories:10K<n<100K",
"source_datasets:extended|other-wider",
"task_categories:object-detection",
"task_ids:face-detection"
] | https://huggingface.co/datasets/coyotte508/dataset/resolve/main/README.md | ---
annotations_creators:
- expert-generated
language_creators:
- found
language:
- en
license:
- cc-by-nc-nd-4.0
multilinguality:
- monolingual
size_categories:
- 10K<n<100K
source_datasets:
- extended|other-wider
task_categories:
- object-detection
task_ids:
- face-detection
paperswithcode_id: wider-face-1
pretty_name: WIDER FACE
---
# Dataset Card for WIDER FACE
## 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:** javascript:alert(document.domain)
- **Repository:**
- **Paper:** [WIDER FACE: A Face Detection Benchmark](https://arxiv.org/abs/1511.06523)
- **Leaderboard:** http://shuoyang1213.me/WIDERFACE/WiderFace_Results.html
- **Point of Contact:** shuoyang.1213@gmail.com
### Dataset Summary
WIDER FACE dataset is a face detection benchmark dataset, of which images are
selected from the publicly available WIDER dataset. We choose 32,203 images and
label 393,703 faces with a high degree of variability in scale, pose and
occlusion as depicted in the sample images. WIDER FACE dataset is organized
based on 61 event classes. For each event class, we randomly select 40%/10%/50%
data as training, validation and testing sets. We adopt the same evaluation
metric employed in the PASCAL VOC dataset. Similar to MALF and Caltech datasets,
we do not release bounding box ground truth for the test images. Users are
required to submit final prediction files, which we shall proceed to evaluate.
### Supported Tasks and Leaderboards
- `face-detection`: The dataset can be used to train a model for Face Detection. More information on evaluating the model's performance can be found [here](http://shuoyang1213.me/WIDERFACE/WiderFace_Results.html).
### Languages
English
## Dataset Structure
### Data Instances
A data point comprises an image and its face annotations.
```
{
'image': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=1024x755 at 0x19FA12186D8>, 'faces': {
'bbox': [
[178.0, 238.0, 55.0, 73.0],
[248.0, 235.0, 59.0, 73.0],
[363.0, 157.0, 59.0, 73.0],
[468.0, 153.0, 53.0, 72.0],
[629.0, 110.0, 56.0, 81.0],
[745.0, 138.0, 55.0, 77.0]
],
'blur': [2, 2, 2, 2, 2, 2],
'expression': [0, 0, 0, 0, 0, 0],
'illumination': [0, 0, 0, 0, 0, 0],
'occlusion': [1, 2, 1, 2, 1, 2],
'pose': [0, 0, 0, 0, 0, 0],
'invalid': [False, False, False, False, False, False]
}
}
```
### Data Fields
- `image`: A `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]`
- `faces`: a dictionary of face attributes for the faces present on the image
- `bbox`: the bounding box of each face (in the [coco](https://albumentations.ai/docs/getting_started/bounding_boxes_augmentation/#coco) format)
- `blur`: the blur level of each face, with possible values including `clear` (0), `normal` (1) and `heavy`
- `expression`: the facial expression of each face, with possible values including `typical` (0) and `exaggerate` (1)
- `illumination`: the lightning condition of each face, with possible values including `normal` (0) and `exaggerate` (1)
- `occlusion`: the level of occlusion of each face, with possible values including `no` (0), `partial` (1) and `heavy` (2)
- `pose`: the pose of each face, with possible values including `typical` (0) and `atypical` (1)
- `invalid`: whether the image is valid or invalid.
### Data Splits
The data is split into training, validation and testing set. WIDER FACE dataset is organized
based on 61 event classes. For each event class, 40%/10%/50%
data is randomly selected as training, validation and testing sets. The training set contains 12880 images, the validation set 3226 images and test set 16097 images.
## Dataset Creation
### Curation Rationale
The curators state that the current face detection datasets typically contain a few thousand faces, with limited variations in pose, scale, facial expression, occlusion, and background clutters,
making it difficult to assess for real world performance. They argue that the limitations of datasets have partially contributed to the failure of some algorithms in coping
with heavy occlusion, small scale, and atypical pose.
### Source Data
#### Initial Data Collection and Normalization
WIDER FACE dataset is a subset of the WIDER dataset.
The images in WIDER were collected in the following three steps: 1) Event categories
were defined and chosen following the Large Scale Ontology for Multimedia (LSCOM) [22], which provides around 1000 concepts relevant to video event analysis. 2) Images
are retrieved using search engines like Google and Bing. For
each category, 1000-3000 images were collected. 3) The
data were cleaned by manually examining all the images
and filtering out images without human face. Then, similar
images in each event category were removed to ensure large
diversity in face appearance. A total of 32203 images are
eventually included in the WIDER FACE dataset.
#### Who are the source language producers?
The images are selected from publicly available WIDER dataset.
### Annotations
#### Annotation process
The curators label the bounding boxes for all
the recognizable faces in the WIDER FACE dataset. The
bounding box is required to tightly contain the forehead,
chin, and cheek.. If a face is occluded, they still label it with a bounding box but with an estimation on the scale of occlusion. Similar to the PASCAL VOC dataset [6], they assign an ’Ignore’ flag to the face
which is very difficult to be recognized due to low resolution and small scale (10 pixels or less). After annotating
the face bounding boxes, they further annotate the following
attributes: pose (typical, atypical) and occlusion level (partial, heavy). Each annotation is labeled by one annotator
and cross-checked by two different people.
#### Who are the annotators?
Shuo Yang, Ping Luo, Chen Change Loy and Xiaoou Tang.
### 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
Shuo Yang, Ping Luo, Chen Change Loy and Xiaoou Tang
### Licensing Information
[Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)](https://creativecommons.org/licenses/by-nc-nd/4.0/).
### Citation Information
```
@inproceedings{yang2016wider,
Author = {Yang, Shuo and Luo, Ping and Loy, Chen Change and Tang, Xiaoou},
Booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
Title = {WIDER FACE: A Face Detection Benchmark},
Year = {2016}}
```
### Contributions
Thanks to [@mariosasko](https://github.com/mariosasko) for adding this dataset. |
csarron | null | null | null | false | 9 | false | csarron/25m-img-caps | 2022-03-28T18:51:26.000Z | null | false | bfd4f4689c343cabfc936eb4c12f026df15cf977 | [] | [] | https://huggingface.co/datasets/csarron/25m-img-caps/resolve/main/README.md | see https://huggingface.co/datasets/csarron/4m-img-caps for example usage |
csarron | null | null | null | false | 10 | false | csarron/4m-img-caps | 2022-03-28T18:50:53.000Z | null | false | b27ebb236e94f8d090891e010f93832dccb034d3 | [] | [] | https://huggingface.co/datasets/csarron/4m-img-caps/resolve/main/README.md | see [read_pyarrow.py](https://gist.github.com/csarron/df712e53c9e0dcaad4eb6843e7a3d51c#file-read_pyarrow-py) for how to read one pyarrow file.
example PyTorch dataset:
```python
from torch.utils.data import Dataset
class ImageCaptionArrowDataset(Dataset):
def __init__(
self,
dataset_file,
tokenizer,
):
import pyarrow as pa
data = [pa.ipc.open_file(pa.memory_map(f, "rb")).read_all() for f in glob.glob(dataset_file)]
self.data = pa.concat_tables(data)
# do other initialization, like init image preprocessing fn,
def __getitem__(self, index):
# item_id = self.data["id"][index].as_py()
text = self.data["text"][index].as_py() # get text
if isinstance(text, list):
text = random.choice(text)
img_bytes = self.data["image"][index].as_py() # get image bytes
# do some processing with image and text, return the features
# img_feat = self.image_bytes_to_tensor(img_bytes)
# inputs = self.tokenizer(
# text,
# padding="max_length",
# max_length=self.max_text_len,
# truncation=True,
# return_token_type_ids=True,
# return_attention_mask=True,
# add_special_tokens=True,
# return_tensors="pt",
# )
# input_ids = inputs.input_ids.squeeze(0)
# attention_mask = inputs.attention_mask.squeeze(0)
# return {
# # "item_ids": item_id,
# "text_ids": input_ids,
# "input_ids": input_ids,
# "text_masks": attention_mask,
# "pixel_values": img_feat,
# }
def __len__(self):
return len(self.data)
``` |
csebuetnlp | null | @inproceedings{hasan-etal-2021-xl,
title = "{XL}-Sum: Large-Scale Multilingual Abstractive Summarization for 44 Languages",
author = "Hasan, Tahmid and
Bhattacharjee, Abhik and
Islam, Md. Saiful and
Mubasshir, Kazi and
Li, Yuan-Fang and
Kang, Yong-Bin and
Rahman, M. Sohel and
Shahriyar, Rifat",
booktitle = "Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021",
month = aug,
year = "2021",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.findings-acl.413",
pages = "4693--4703",
} | We present XLSum, a comprehensive and diverse dataset comprising 1.35 million professionally
annotated article-summary pairs from BBC, extracted using a set of carefully designed heuristics.
The dataset covers 45 languages ranging from low to high-resource, for many of which no
public dataset is currently available. XL-Sum is highly abstractive, concise,
and of high quality, as indicated by human and intrinsic evaluation. | false | 7,806 | false | csebuetnlp/xlsum | 2022-08-10T11:33:03.000Z | xl-sum | false | 33bf2120fc639aac8c9ebc3248d77618efb9d7d6 | [] | [
"arxiv:1607.01759",
"task_ids:summarization",
"language:am",
"language:ar",
"language:az",
"language:bn",
"language:my",
"language:zh",
"language:en",
"language:fr",
"language:gu",
"language:ha",
"language:hi",
"language:ig",
"language:id",
"language:ja",
"language:rn",
"language:ko",
"language:ky",
"language:mr",
"language:ne",
"language:om",
"language:ps",
"language:fa",
"language:pcm",
"language:pt",
"language:pa",
"language:ru",
"language:gd",
"language:sr",
"language:si",
"language:so",
"language:es",
"language:sw",
"language:ta",
"language:te",
"language:th",
"language:ti",
"language:tr",
"language:uk",
"language:ur",
"language:uz",
"language:vi",
"language:cy",
"language:yo",
"size_categories:1M<n<10M",
"license:cc-by-nc-sa-4.0",
"multilinguality:multilingual",
"source_datasets:original",
"annotations_creators:found",
"language_creators:found"
] | https://huggingface.co/datasets/csebuetnlp/xlsum/resolve/main/README.md | ---
task_categories:
- conditional-text-generation
task_ids:
- summarization
language:
- am
- ar
- az
- bn
- my
- zh
- en
- fr
- gu
- ha
- hi
- ig
- id
- ja
- rn
- ko
- ky
- mr
- ne
- om
- ps
- fa
- pcm
- pt
- pa
- ru
- gd
- sr
- si
- so
- es
- sw
- ta
- te
- th
- ti
- tr
- uk
- ur
- uz
- vi
- cy
- yo
size_categories:
- 1M<n<10M
license:
- cc-by-nc-sa-4.0
multilinguality:
- multilingual
source_datasets:
- original
paperswithcode_id: xl-sum
annotations_creators:
- found
language_creators:
- found
pretty_name: XL-Sum
---
# Dataset Card for "XL-Sum"
## Table of Contents
- [Dataset Card Creation Guide](#dataset-card-creation-guide)
- [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
- **Repository:** [https://github.com/csebuetnlp/xl-sum](https://github.com/csebuetnlp/xl-sum)
- **Paper:** [XL-Sum: Large-Scale Multilingual Abstractive Summarization for 44 Languages](https://aclanthology.org/2021.findings-acl.413/)
- **Point of Contact:** [Tahmid Hasan](mailto:tahmidhasan@cse.buet.ac.bd)
### Dataset Summary
We present XLSum, a comprehensive and diverse dataset comprising 1.35 million professionally annotated article-summary pairs from BBC, extracted using a set of carefully designed heuristics. The dataset covers 45 languages ranging from low to high-resource, for many of which no public dataset is currently available. XL-Sum is highly abstractive, concise, and of high quality, as indicated by human and intrinsic evaluation.
### Supported Tasks and Leaderboards
[More information needed](https://github.com/csebuetnlp/xl-sum)
### Languages
- `amharic`
- `arabic`
- `azerbaijani`
- `bengali`
- `burmese`
- `chinese_simplified`
- `chinese_traditional`
- `english`
- `french`
- `gujarati`
- `hausa`
- `hindi`
- `igbo`
- `indonesian`
- `japanese`
- `kirundi`
- `korean`
- `kyrgyz`
- `marathi`
- `nepali`
- `oromo`
- `pashto`
- `persian`
- `pidgin`
- `portuguese`
- `punjabi`
- `russian`
- `scottish_gaelic`
- `serbian_cyrillic`
- `serbian_latin`
- `sinhala`
- `somali`
- `spanish`
- `swahili`
- `tamil`
- `telugu`
- `thai`
- `tigrinya`
- `turkish`
- `ukrainian`
- `urdu`
- `uzbek`
- `vietnamese`
- `welsh`
- `yoruba`
## Dataset Structure
### Data Instances
One example from the `English` dataset is given below in JSON format.
```
{
"id": "technology-17657859",
"url": "https://www.bbc.com/news/technology-17657859",
"title": "Yahoo files e-book advert system patent applications",
"summary": "Yahoo has signalled it is investigating e-book adverts as a way to stimulate its earnings.",
"text": "Yahoo's patents suggest users could weigh the type of ads against the sizes of discount before purchase. It says in two US patent applications that ads for digital book readers have been \"less than optimal\" to date. The filings suggest that users could be offered titles at a variety of prices depending on the ads' prominence They add that the products shown could be determined by the type of book being read, or even the contents of a specific chapter, phrase or word. The paperwork was published by the US Patent and Trademark Office late last week and relates to work carried out at the firm's headquarters in Sunnyvale, California. \"Greater levels of advertising, which may be more valuable to an advertiser and potentially more distracting to an e-book reader, may warrant higher discounts,\" it states. Free books It suggests users could be offered ads as hyperlinks based within the book's text, in-laid text or even \"dynamic content\" such as video. Another idea suggests boxes at the bottom of a page could trail later chapters or quotes saying \"brought to you by Company A\". It adds that the more willing the customer is to see the ads, the greater the potential discount. \"Higher frequencies... may even be great enough to allow the e-book to be obtained for free,\" it states. The authors write that the type of ad could influence the value of the discount, with \"lower class advertising... such as teeth whitener advertisements\" offering a cheaper price than \"high\" or \"middle class\" adverts, for things like pizza. The inventors also suggest that ads could be linked to the mood or emotional state the reader is in as a they progress through a title. For example, they say if characters fall in love or show affection during a chapter, then ads for flowers or entertainment could be triggered. The patents also suggest this could applied to children's books - giving the Tom Hanks animated film Polar Express as an example. It says a scene showing a waiter giving the protagonists hot drinks \"may be an excellent opportunity to show an advertisement for hot cocoa, or a branded chocolate bar\". Another example states: \"If the setting includes young characters, a Coke advertisement could be provided, inviting the reader to enjoy a glass of Coke with his book, and providing a graphic of a cool glass.\" It adds that such targeting could be further enhanced by taking account of previous titles the owner has bought. 'Advertising-free zone' At present, several Amazon and Kobo e-book readers offer full-screen adverts when the device is switched off and show smaller ads on their menu screens, but the main text of the titles remains free of marketing. Yahoo does not currently provide ads to these devices, and a move into the area could boost its shrinking revenues. However, Philip Jones, deputy editor of the Bookseller magazine, said that the internet firm might struggle to get some of its ideas adopted. \"This has been mooted before and was fairly well decried,\" he said. \"Perhaps in a limited context it could work if the merchandise was strongly related to the title and was kept away from the text. \"But readers - particularly parents - like the fact that reading is an advertising-free zone. Authors would also want something to say about ads interrupting their narrative flow.\""
}
```
### Data Fields
- 'id': A string representing the article ID.
- 'url': A string representing the article URL.
- 'title': A string containing the article title.
- 'summary': A string containing the article summary.
- 'text' : A string containing the article text.
### Data Splits
We used a 80%-10%-10% split for all languages with a few exceptions. `English` was split 93%-3.5%-3.5% for the evaluation set size to resemble that of `CNN/DM` and `XSum`; `Scottish Gaelic`, `Kyrgyz` and `Sinhala` had relatively fewer samples, their evaluation sets were increased to 500 samples for more reliable evaluation. Same articles were used for evaluation in the two variants of Chinese and Serbian to prevent data leakage in multilingual training. Individual dataset download links with train-dev-test example counts are given below:
Language | ISO 639-1 Code | BBC subdomain(s) | Train | Dev | Test | Total |
--------------|----------------|------------------|-------|-----|------|-------|
Amharic | am | https://www.bbc.com/amharic | 5761 | 719 | 719 | 7199 |
Arabic | ar | https://www.bbc.com/arabic | 37519 | 4689 | 4689 | 46897 |
Azerbaijani | az | https://www.bbc.com/azeri | 6478 | 809 | 809 | 8096 |
Bengali | bn | https://www.bbc.com/bengali | 8102 | 1012 | 1012 | 10126 |
Burmese | my | https://www.bbc.com/burmese | 4569 | 570 | 570 | 5709 |
Chinese (Simplified) | zh-CN | https://www.bbc.com/ukchina/simp, https://www.bbc.com/zhongwen/simp | 37362 | 4670 | 4670 | 46702 |
Chinese (Traditional) | zh-TW | https://www.bbc.com/ukchina/trad, https://www.bbc.com/zhongwen/trad | 37373 | 4670 | 4670 | 46713 |
English | en | https://www.bbc.com/english, https://www.bbc.com/sinhala `*` | 306522 | 11535 | 11535 | 329592 |
French | fr | https://www.bbc.com/afrique | 8697 | 1086 | 1086 | 10869 |
Gujarati | gu | https://www.bbc.com/gujarati | 9119 | 1139 | 1139 | 11397 |
Hausa | ha | https://www.bbc.com/hausa | 6418 | 802 | 802 | 8022 |
Hindi | hi | https://www.bbc.com/hindi | 70778 | 8847 | 8847 | 88472 |
Igbo | ig | https://www.bbc.com/igbo | 4183 | 522 | 522 | 5227 |
Indonesian | id | https://www.bbc.com/indonesia | 38242 | 4780 | 4780 | 47802 |
Japanese | ja | https://www.bbc.com/japanese | 7113 | 889 | 889 | 8891 |
Kirundi | rn | https://www.bbc.com/gahuza | 5746 | 718 | 718 | 7182 |
Korean | ko | https://www.bbc.com/korean | 4407 | 550 | 550 | 5507 |
Kyrgyz | ky | https://www.bbc.com/kyrgyz | 2266 | 500 | 500 | 3266 |
Marathi | mr | https://www.bbc.com/marathi | 10903 | 1362 | 1362 | 13627 |
Nepali | np | https://www.bbc.com/nepali | 5808 | 725 | 725 | 7258 |
Oromo | om | https://www.bbc.com/afaanoromoo | 6063 | 757 | 757 | 7577 |
Pashto | ps | https://www.bbc.com/pashto | 14353 | 1794 | 1794 | 17941 |
Persian | fa | https://www.bbc.com/persian | 47251 | 5906 | 5906 | 59063 |
Pidgin`**` | n/a | https://www.bbc.com/pidgin | 9208 | 1151 | 1151 | 11510 |
Portuguese | pt | https://www.bbc.com/portuguese | 57402 | 7175 | 7175 | 71752 |
Punjabi | pa | https://www.bbc.com/punjabi | 8215 | 1026 | 1026 | 10267 |
Russian | ru | https://www.bbc.com/russian, https://www.bbc.com/ukrainian `*` | 62243 | 7780 | 7780 | 77803 |
Scottish Gaelic | gd | https://www.bbc.com/naidheachdan | 1313 | 500 | 500 | 2313 |
Serbian (Cyrillic) | sr | https://www.bbc.com/serbian/cyr | 7275 | 909 | 909 | 9093 |
Serbian (Latin) | sr | https://www.bbc.com/serbian/lat | 7276 | 909 | 909 | 9094 |
Sinhala | si | https://www.bbc.com/sinhala | 3249 | 500 | 500 | 4249 |
Somali | so | https://www.bbc.com/somali | 5962 | 745 | 745 | 7452 |
Spanish | es | https://www.bbc.com/mundo | 38110 | 4763 | 4763 | 47636 |
Swahili | sw | https://www.bbc.com/swahili | 7898 | 987 | 987 | 9872 |
Tamil | ta | https://www.bbc.com/tamil | 16222 | 2027 | 2027 | 20276 |
Telugu | te | https://www.bbc.com/telugu | 10421 | 1302 | 1302 | 13025 |
Thai | th | https://www.bbc.com/thai | 6616 | 826 | 826 | 8268 |
Tigrinya | ti | https://www.bbc.com/tigrinya | 5451 | 681 | 681 | 6813 |
Turkish | tr | https://www.bbc.com/turkce | 27176 | 3397 | 3397 | 33970 |
Ukrainian | uk | https://www.bbc.com/ukrainian | 43201 | 5399 | 5399 | 53999 |
Urdu | ur | https://www.bbc.com/urdu | 67665 | 8458 | 8458 | 84581 |
Uzbek | uz | https://www.bbc.com/uzbek | 4728 | 590 | 590 | 5908 |
Vietnamese | vi | https://www.bbc.com/vietnamese | 32111 | 4013 | 4013 | 40137 |
Welsh | cy | https://www.bbc.com/cymrufyw | 9732 | 1216 | 1216 | 12164 |
Yoruba | yo | https://www.bbc.com/yoruba | 6350 | 793 | 793 | 7936 |
`*` A lot of articles in BBC Sinhala and BBC Ukrainian were written in English and Russian respectively. They were identified using [Fasttext](https://arxiv.org/abs/1607.01759) and moved accordingly.
`**` West African Pidgin English
## Dataset Creation
### Curation Rationale
[More information needed](https://github.com/csebuetnlp/xl-sum)
### Source Data
[BBC News](https://www.bbc.co.uk/ws/languages)
#### Initial Data Collection and Normalization
[Detailed in the paper](https://aclanthology.org/2021.findings-acl.413/)
#### Who are the source language producers?
[Detailed in the paper](https://aclanthology.org/2021.findings-acl.413/)
### Annotations
[Detailed in the paper](https://aclanthology.org/2021.findings-acl.413/)
#### Annotation process
[Detailed in the paper](https://aclanthology.org/2021.findings-acl.413/)
#### Who are the annotators?
[Detailed in the paper](https://aclanthology.org/2021.findings-acl.413/)
### Personal and Sensitive Information
[More information needed](https://github.com/csebuetnlp/xl-sum)
## Considerations for Using the Data
### Social Impact of Dataset
[More information needed](https://github.com/csebuetnlp/xl-sum)
### Discussion of Biases
[More information needed](https://github.com/csebuetnlp/xl-sum)
### Other Known Limitations
[More information needed](https://github.com/csebuetnlp/xl-sum)
## Additional Information
### Dataset Curators
[More information needed](https://github.com/csebuetnlp/xl-sum)
### Licensing Information
Contents of this repository are restricted to only non-commercial research purposes under the [Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0)](https://creativecommons.org/licenses/by-nc-sa/4.0/). Copyright of the dataset contents belongs to the original copyright holders.
### Citation Information
If you use any of the datasets, models or code modules, please cite the following paper:
```
@inproceedings{hasan-etal-2021-xl,
title = "{XL}-Sum: Large-Scale Multilingual Abstractive Summarization for 44 Languages",
author = "Hasan, Tahmid and
Bhattacharjee, Abhik and
Islam, Md. Saiful and
Mubasshir, Kazi and
Li, Yuan-Fang and
Kang, Yong-Bin and
Rahman, M. Sohel and
Shahriyar, Rifat",
booktitle = "Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021",
month = aug,
year = "2021",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.findings-acl.413",
pages = "4693--4703",
}
```
### Contributions
Thanks to [@abhik1505040](https://github.com/abhik1505040) and [@Tahmid](https://github.com/Tahmid04) for adding this dataset. |
csebuetnlp | null | @misc{bhattacharjee2021banglabert,
title={BanglaBERT: Combating Embedding Barrier in Multilingual Models for Low-Resource Language Understanding},
author={Abhik Bhattacharjee and Tahmid Hasan and Kazi Samin and Md Saiful Islam and M. Sohel Rahman and Anindya Iqbal and Rifat Shahriyar},
year={2021},
eprint={2101.00204},
archivePrefix={arXiv},
primaryClass={cs.CL}
} | This is a Natural Language Inference (NLI) dataset for Bengali, curated using the subset of
MNLI data used in XNLI and state-of-the-art English to Bengali translation model. | false | 339 | false | csebuetnlp/xnli_bn | 2022-08-21T13:14:56.000Z | null | false | a18ecb62d7ffd4a6bff5756afb6e799bbb91dd3e | [] | [
"arxiv:2101.00204",
"arxiv:2007.01852",
"annotations_creators:machine-generated",
"language_creators:found",
"multilinguality:monolingual",
"size_categories:100K<n<1M",
"source_datasets:extended",
"task_categories:text-classification",
"task_ids:natural-language-inference",
"language:bn",
"license:cc-by-nc-sa-4.0"
] | https://huggingface.co/datasets/csebuetnlp/xnli_bn/resolve/main/README.md | ---
annotations_creators:
- machine-generated
language_creators:
- found
multilinguality:
- monolingual
size_categories:
- 100K<n<1M
source_datasets:
- extended
task_categories:
- text-classification
task_ids:
- natural-language-inference
language:
- bn
license:
- cc-by-nc-sa-4.0
---
# Dataset Card for `xnli_bn`
## Table of Contents
- [Dataset Card for `xnli_bn`](#dataset-card-for-xnli_bn)
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Dataset Summary](#dataset-summary)
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
- [Languages](#languages)
- [Usage](#usage)
- [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
- **Repository:** [https://github.com/csebuetnlp/banglabert](https://github.com/csebuetnlp/banglabert)
- **Paper:** [**"BanglaBERT: Combating Embedding Barrier in Multilingual Models for Low-Resource Language Understanding"**](https://arxiv.org/abs/2101.00204)
- **Point of Contact:** [Tahmid Hasan](mailto:tahmidhasan@cse.buet.ac.bd)
### Dataset Summary
This is a Natural Language Inference (NLI) dataset for Bengali, curated using the subset of
MNLI data used in XNLI and state-of-the-art English to Bengali translation model introduced **[here](https://aclanthology.org/2020.emnlp-main.207/).**
### Supported Tasks and Leaderboards
[More information needed](https://github.com/csebuetnlp/banglabert)
### Languages
* `Bengali`
### Usage
```python
from datasets import load_dataset
dataset = load_dataset("csebuetnlp/xnli_bn")
```
## Dataset Structure
### Data Instances
One example from the dataset is given below in JSON format.
```
{
"sentence1": "আসলে, আমি এমনকি এই বিষয়ে চিন্তাও করিনি, কিন্তু আমি এত হতাশ হয়ে পড়েছিলাম যে, শেষ পর্যন্ত আমি আবার তার সঙ্গে কথা বলতে শুরু করেছিলাম",
"sentence2": "আমি তার সাথে আবার কথা বলিনি।",
"label": "contradiction"
}
```
### Data Fields
The data fields are as follows:
- `sentence1`: a `string` feature indicating the premise.
- `sentence2`: a `string` feature indicating the hypothesis.
- `label`: a classification label, where possible values are `contradiction` (0), `entailment` (1), `neutral` (2) .
### Data Splits
| split |count |
|----------|--------|
|`train`| 381449 |
|`validation`| 2419 |
|`test`| 4895 |
## Dataset Creation
The dataset curation procedure was the same as the [XNLI](https://aclanthology.org/D18-1269/) dataset: we translated the [MultiNLI](https://aclanthology.org/N18-1101/) training data using the English to Bangla translation model introduced [here](https://aclanthology.org/2020.emnlp-main.207/). Due to the possibility of incursions of error during automatic translation, we used the [Language-Agnostic BERT Sentence Embeddings (LaBSE)](https://arxiv.org/abs/2007.01852) of the translations and original sentences to compute their similarity. All sentences below a similarity threshold of 0.70 were discarded.
### Curation Rationale
[More information needed](https://github.com/csebuetnlp/banglabert)
### Source Data
[XNLI](https://aclanthology.org/D18-1269/)
#### Initial Data Collection and Normalization
[More information needed](https://github.com/csebuetnlp/banglabert)
#### Who are the source language producers?
[More information needed](https://github.com/csebuetnlp/banglabert)
### Annotations
[More information needed](https://github.com/csebuetnlp/banglabert)
#### Annotation process
[More information needed](https://github.com/csebuetnlp/banglabert)
#### Who are the annotators?
[More information needed](https://github.com/csebuetnlp/banglabert)
### Personal and Sensitive Information
[More information needed](https://github.com/csebuetnlp/banglabert)
## Considerations for Using the Data
### Social Impact of Dataset
[More information needed](https://github.com/csebuetnlp/banglabert)
### Discussion of Biases
[More information needed](https://github.com/csebuetnlp/banglabert)
### Other Known Limitations
[More information needed](https://github.com/csebuetnlp/banglabert)
## Additional Information
### Dataset Curators
[More information needed](https://github.com/csebuetnlp/banglabert)
### Licensing Information
Contents of this repository are restricted to only non-commercial research purposes under the [Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0)](https://creativecommons.org/licenses/by-nc-sa/4.0/). Copyright of the dataset contents belongs to the original copyright holders.
### Citation Information
If you use the dataset, please cite the following paper:
```
@misc{bhattacharjee2021banglabert,
title={BanglaBERT: Combating Embedding Barrier in Multilingual Models for Low-Resource Language Understanding},
author={Abhik Bhattacharjee and Tahmid Hasan and Kazi Samin and Md Saiful Islam and M. Sohel Rahman and Anindya Iqbal and Rifat Shahriyar},
year={2021},
eprint={2101.00204},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
```
### Contributions
Thanks to [@abhik1505040](https://github.com/abhik1505040) and [@Tahmid](https://github.com/Tahmid04) for adding this dataset. |
cstrathe435 | null | null | null | false | 324 | false | cstrathe435/Task2Dial | 2022-02-03T12:55:28.000Z | null | false | d810e76b4b49ceffb417666524b0daabd94c059c | [] | [] | https://huggingface.co/datasets/cstrathe435/Task2Dial/resolve/main/README.md | # Dataset Card for Task2Dial
## 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)
- [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)
- [Acknowledgements] (#funding-information)
## Dataset Description
- **Homepage:** [Needs More Information]
- **Repository:** [Needs More Information]
- **Paper:** https://aclanthology.org/2021.icnlsp-1.28/
- **Leaderboard:** [Needs More Information]
- **Point of Contact:** [Needs More Information]
### Dataset Summary
The Task2Dial dataset includes (1) a set of recipe documents with 353 individual dialogues; and (2) conversations between an IG and an IF, which are grounded in the associated recipe documents. Presents sample utterances from a dialogue along with the associated recipe. It demonstrates some important features of the dataset, such as mentioning entities not present in the recipe document; re-composition of the original text to focus on the important steps and the breakdown of the recipe into manageable and appropriate steps. Following recent efforts in the field to standardise NLG research, we have made the dataset freely available.
### Supported Tasks and Leaderboards
We demonstrate the task of implementing the Task2Dial in a conversational agent called chefbot in the following git repo: https://github.com/carlstrath/ChefBot
### Languages
English
### Data Fields
Dataset.1: Task2Dial main, 353 cooking recipes modelled on real conversations between an IF and IG.
Dataset. 2: A list of alternative ingredients for every swappable ingredient in the Task2Dial dataset.
Dataset. 3. A list of objects and utensils with explanations, comparisons, handling and common storage location information.
## Dataset Creation
The proposed task considers the recipe-following scenario with an information giver
(IG) and an information follower (IF), where the IG has access to the recipe and gives
instructions to the IF. The IG might choose to omit irrelevant information, simplify
the content of a recipe or provide it as is. The IF will either follow the task or ask
for further information. The IG might have to rely on information outside the given
document (i.e. commonsense) to enhance understanding and success of the task. In
addition, the IG decides on how to present the recipe steps, i.e. split them into sub-
steps or merge them together, often diverging from the original number of recipe steps.
The task is regarded as successful when the IG has successfully followed/understood
the recipe. Hence, other dialogue-focused metrics, such as the number of turns, are
not appropriate here. Formally, Task2Dial can be defined as follows: Given a recipe
𝑅𝑖 from 𝑅 =𝑅1, 𝑅2, 𝑅3,..., 𝑅𝑛, an ontology or ontologies 𝑂𝑖 =𝑂11,𝑂2,...,𝑂𝑛 of
cooking-related concepts, a history of the conversation ℎ, predict the response 𝑟 of
the IG.
### Curation Rationale
Text selection was dependent on the quality of the information
provided in the existing recipes. Too little information and the transcription and
interpretation of the text became diffused with missing or incorrect knowledge.
Conversely, providing too much information in the text resulted in a lack of creativity
and commonsense reasoning by the data curators. Thus, the goal of the curation was
to identify text that contained all the relevant information to complete the cooking
task (tools, ingredients, weights, timings, servings) but not in such detail that it
subtracted from the creativity, commonsense and imagination of the annotators.
### Source Data
#### Initial Data Collection and Normalization
Three open-source and creative commons licensed
cookery websites6 were identified for data extraction, which permits any use or non-
commercial use of data for research purposes. As content submission to the
cooking websites was unrestricted, data appropriateness was ratified by the ratings
and reviews given to each recipe by the public, highly rated recipes with a positive
feedback were given preference over recipes with low scores and poor reviews [38].
From this, a list of 353 recipes was compiled and divided amongst the annotators
for the data collection. As mentioned earlier, annotators were asked to take on the
roles of both IF and IG, rather than a multi-turn WoZ approach, to allow flexibility
in the utterances. This approach allowed the annotators additional time to formulate
detailed and concise responses.
#### Who are the source language producers?
Undergraduate RAs were recruited through email.
The participants were paid an hourly rate based on a university pay scale which is
above the living wage and corresponds to the real living wage, following ethical
guidelines for responsible innovation. The annotation team was composed of
two males and one female data curators, under the age of 25 of mixed ethnicity’s with
experience in AI and computing. This minimised the gender bias that is frequently
observed in crowdsourcing platforms.
#### Annotation process
Each annotator was provided with a detailed list of instructions, an example dialogue and an IF/IG template (see Appendix A). The annotators were asked to read both the example dialogue and the original recipe to understand the text, context, composition, translation and annotation. The instructions included information handling and storage of data, text formatting, metadata and examples of high-quality and poor dialogues. An administrator was on hand throughout the data collection to support and guide the annotators. This approach reduced the number of low-quality dialogues associated with large crowdsourcing platforms that are often discarded post evaluation, as demonstrated in the data collection of the Doc2Dial dataset.
#### Who are the annotators?
Research assistants (RAs) from the School of Computing were employed on temporary contracts to construct and format the dataset. After an initial meeting to discuss the job role and determine suitability, the RAs were asked to complete a paid trial, this was evaluated and further advice was given on how to write dialogues and format the data to ensure high quality. After the successful completion of the trial, the RAs were permitted to continue with the remainder of the data collection. To ensure the high quality of the dataset, samples of the dialogues were often reviewed and further feedback was provided.
### Personal and Sensitive Information
An ethics request was submitted for review by the board of ethics at our university. No personal or other data that may be used to identify an individual was collected in this study.
## Considerations for Using the Data
The Task2Dial dataset is currently only for the cooking domain, but using the methodologies provided other tasks can be modelled for example, furniture assembly and maintenance tasks.
### Social Impact of Dataset
Our proposed task aims to motivate research for modern dialogue systems that
address the following challenges. Firstly, modern dialogue systems should be flexible
and allow for "off-script" scenarios in order to emulate real-world phenomena, such
as the ones present in human-human communication. This will require new ways
of encoding user intents and new approaches to dialogue management in general.
Secondly, as dialogue systems find different domain applications, the complexity
of the dialogues might increase as well as the reliance on domain knowledge that
can be encoded in structured or unstructured ways, such as documents, databases
etc. Many applications, might require access to different domain knowledge sources
in a course of a dialogue, and in such context, selection might prove beneficial in
choosing "what to say".
### Discussion of Biases
Prior to data collection, we performed three pilot studies.
In the first, two participants assumed the roles of IG and IF respectively, where the
IG had access to a recipe and provided recipe instructions to the IF (who did not have
access to the recipe) over the phone, recording the session and then transcribing it.
Next, we repeated the process with text-based dialogue through an online platform
following a similar setup, however, the interaction was solely chat-based. The final
study used self-dialogue, with one member of the team writing entire dialogues
assuming both the IF and IG roles. We found that self-dialogue results were proximal
to the results of two-person studies. However, time and cost were higher for producing
two-person dialogues, with the additional time needed for transcribing and correction,
thus, we opted to use self-dialogue.
## Additional Information
Video: https://www.youtube.com/watch?v=zISkwn95RXs&ab_channel=ICNLSPConference
### Dataset Curators
The recipes are composed by people of a different races
/ ethnicity, nationalities, socioeconomic status, abilities, age, gender and language
with significant variation in pronunciations, structure, language and grammar. This
provided the annotators with unique linguistic content for each recipe to interpret
the data and configure the text into an IF/IG format. To help preserve sociolinguistic
patterns in speech, the data curators retained the underlying language when para-
phrasing, to intercede social and regional dialects with their own interpretation of
the data to enhance the lexical richness.
### Licensing Information
CC
### Citation Information
https://aclanthology.org/2021.icnlsp-1.28/
### Acknowledgements
The research is supported under the EPSRC projects CiViL (EP/T014598/1) and
NLG for low-resource domains (EP/T024917/1). |
ctu-aic | null | @article{DBLP:journals/corr/abs-2201-11115,
author = {Jan Drchal and
Herbert Ullrich and
Martin R{\'{y}}par and
Hana Vincourov{\'{a}} and
V{\'{a}}clav Moravec},
title = {CsFEVER and CTKFacts: Czech Datasets for Fact Verification},
journal = {CoRR},
volume = {abs/2201.11115},
year = {2022},
url = {https://arxiv.org/abs/2201.11115},
eprinttype = {arXiv},
eprint = {2201.11115},
timestamp = {Tue, 01 Feb 2022 14:59:01 +0100},
biburl = {https://dblp.org/rec/journals/corr/abs-2201-11115.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
} | CsFEVER is a Czech localisation of the English FEVER datgaset. | false | 321 | false | ctu-aic/csfever | 2022-11-01T05:56:15.000Z | null | false | c9f2ce78fc92e19353b7f1cb3f4b68f15d32eb1c | [] | [
"arxiv:1803.05355",
"arxiv:2201.11115",
"license:cc-by-sa-3.0"
] | https://huggingface.co/datasets/ctu-aic/csfever/resolve/main/README.md | ---
license: cc-by-sa-3.0
---
# CsFEVER experimental Fact-Checking dataset
Czech dataset for fact verification localized from the data points of [FEVER](https://arxiv.org/abs/1803.05355) using the localization scheme described in the [CTKFacts: Czech Datasets for Fact Verification](https://arxiv.org/abs/2201.11115) paper which is currently being revised for publication in LREV journal.
The version you are looking at was reformatted to *Claim*-*Evidence* string pairs for the specific task of NLI - a more general Document-Retrieval-ready interpretation of our datapoints which can be used for training and evaluating the DR models over the June 2016 wikipedia snapshot can be found in the [data_dr]() folder in the JSON Lines format.
## Data Statement
### Curation Rationale
TODO
|
ctu-aic | null | todo | CsfeverNLI is a NLI version of the Czech Csfever dataset | false | 321 | false | ctu-aic/csfever_nli | 2022-02-22T11:13:35.000Z | null | false | 69d0247380ab01c39f2920974a1736e92fe45783 | [] | [] | https://huggingface.co/datasets/ctu-aic/csfever_nli/resolve/main/README.md | |
ctu-aic | null | @article{DBLP:journals/corr/abs-2201-11115,
author = {Jan Drchal and
Herbert Ullrich and
Martin R{\'{y}}par and
Hana Vincourov{\'{a}} and
V{\'{a}}clav Moravec},
title = {CsFEVER and CTKFacts: Czech Datasets for Fact Verification},
journal = {CoRR},
volume = {abs/2201.11115},
year = {2022},
url = {https://arxiv.org/abs/2201.11115},
eprinttype = {arXiv},
eprint = {2201.11115},
timestamp = {Tue, 01 Feb 2022 14:59:01 +0100},
biburl = {https://dblp.org/rec/journals/corr/abs-2201-11115.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
} | CtkFactsNLI is a NLI version of the Czech CTKFacts dataset | false | 322 | false | ctu-aic/ctkfacts_nli | 2022-11-01T06:35:47.000Z | null | false | 387ae4582c8054cb52ef57ef0941f19bd8012abf | [] | [
"arxiv:2201.11115"
] | https://huggingface.co/datasets/ctu-aic/ctkfacts_nli/resolve/main/README.md | # CTKFacts dataset for Natural Language Inference
Czech Natural Language Inference dataset of ~3K *evidence*-*claim* pairs labelled with SUPPORTS, REFUTES or NOT ENOUGH INFO veracity labels. Extracted from a round of fact-checking experiments concluded and described within the CsFEVER and [CTKFacts: Czech Datasets for Fact Verification](https://arxiv.org/abs/2201.11115) paper currently being revised for publication in LREV journal.
## Document retrieval version
Can be found at https://huggingface.co/datasets/ctu-aic/ctkfacts |
cylee | null | null | null | false | 322 | false | cylee/github-issues | 2021-12-19T19:12:55.000Z | null | false | 3768a20ee7e29288ea5feb4531fc5ab68ca8c2f2 | [] | [
"arxiv:2005.00614"
] | https://huggingface.co/datasets/cylee/github-issues/resolve/main/README.md | # Dataset Card for GitHub Issues
## Dataset Description
This dataset is created for the Hugging Face Datasets library course
### Dataset Summary
GitHub Issues is a dataset consisting of GitHub issues and pull requests associated with the 🤗 Datasets [repository](https://github.com/huggingface/datasets). It is intended for educational purposes and can be used for semantic search or multilabel text classification. The contents of each GitHub issue are in English and concern the domain of datasets for NLP, computer vision, and beyond.
### Supported Tasks and Leaderboards
For each of the tasks tagged for this dataset, give a brief description of the tag, metrics, and suggested models (with a link to their HuggingFace implementation if available). Give a similar description of tasks that were not covered by the structured tag set (repace the `task-category-tag` with an appropriate `other:other-task-name`).
- `task-category-tag`: The dataset can be used to train a model for [TASK NAME], which consists in [TASK DESCRIPTION]. Success on this task is typically measured by achieving a *high/low* [metric name](https://huggingface.co/metrics/metric_name). The ([model name](https://huggingface.co/model_name) or [model class](https://huggingface.co/transformers/model_doc/model_class.html)) model currently achieves the following score. *[IF A LEADERBOARD IS AVAILABLE]:* This task has an active leaderboard which can be found at [leaderboard url]() and ranks models based on [metric name](https://huggingface.co/metrics/metric_name) while also reporting [other metric name](https://huggingface.co/metrics/other_metric_name).
### Languages
Provide a brief overview of the languages represented in the dataset. Describe relevant details about specifics of the language such as whether it is social media text, African American English,...
When relevant, please provide [BCP-47 codes](https://tools.ietf.org/html/bcp47), which consist of a [primary language subtag](https://tools.ietf.org/html/bcp47#section-2.2.1), with a [script subtag](https://tools.ietf.org/html/bcp47#section-2.2.3) and/or [region subtag](https://tools.ietf.org/html/bcp47#section-2.2.4) if available.
## Dataset Structure
### Data Instances
Provide an JSON-formatted example and brief description of a typical instance in the dataset. If available, provide a link to further examples.
```
{
'example_field': ...,
...
}
```
Provide any additional information that is not covered in the other sections about the data here. In particular describe any relationships between data points and if these relationships are made explicit.
### Data Fields
List and describe the fields present in the dataset. Mention their data type, and whether they are used as input or output in any of the tasks the dataset currently supports. If the data has span indices, describe their attributes, such as whether they are at the character level or word level, whether they are contiguous or not, etc. If the datasets contains example IDs, state whether they have an inherent meaning, such as a mapping to other datasets or pointing to relationships between data points.
- `example_field`: description of `example_field`
Note that the descriptions can be initialized with the **Show Markdown Data Fields** output of the [tagging app](https://github.com/huggingface/datasets-tagging), you will then only need to refine the generated descriptions.
### Data Splits
Describe and name the splits in the dataset if there are more than one.
Describe any criteria for splitting the data, if used. If their are differences between the splits (e.g. if the training annotations are machine-generated and the dev and test ones are created by humans, or if different numbers of annotators contributed to each example), describe them here.
Provide the sizes of each split. As appropriate, provide any descriptive statistics for the features, such as average length. For example:
| | Tain | Valid | Test |
| ----- | ------ | ----- | ---- |
| Input Sentences | | | |
| Average Sentence Length | | | |
## Dataset Creation
### Curation Rationale
What need motivated the creation of this dataset? What are some of the reasons underlying the major choices involved in putting it together?
### Source Data
This section describes the source data (e.g. news text and headlines, social media posts, translated sentences,...)
#### Initial Data Collection and Normalization
Describe the data collection process. Describe any criteria for data selection or filtering. List any key words or search terms used. If possible, include runtime information for the collection process.
If data was collected from other pre-existing datasets, link to source here and to their [Hugging Face version](https://huggingface.co/datasets/dataset_name).
If the data was modified or normalized after being collected (e.g. if the data is word-tokenized), describe the process and the tools used.
#### Who are the source language producers?
State whether the data was produced by humans or machine generated. Describe the people or systems who originally created the data.
If available, include self-reported demographic or identity information for the source data creators, but avoid inferring this information. Instead state that this information is unknown. See [Larson 2017](https://www.aclweb.org/anthology/W17-1601.pdf) for using identity categories as a variables, particularly gender.
Describe the conditions under which the data was created (for example, if the producers were crowdworkers, state what platform was used, or if the data was found, what website the data was found on). If compensation was provided, include that information here.
Describe other people represented or mentioned in the data. Where possible, link to references for the information.
### Annotations
If the dataset contains annotations which are not part of the initial data collection, describe them in the following paragraphs.
#### Annotation process
If applicable, describe the annotation process and any tools used, or state otherwise. Describe the amount of data annotated, if not all. Describe or reference annotation guidelines provided to the annotators. If available, provide interannotator statistics. Describe any annotation validation processes.
#### Who are the annotators?
If annotations were collected for the source data (such as class labels or syntactic parses), state whether the annotations were produced by humans or machine generated.
Describe the people or systems who originally created the annotations and their selection criteria if applicable.
If available, include self-reported demographic or identity information for the annotators, but avoid inferring this information. Instead state that this information is unknown. See [Larson 2017](https://www.aclweb.org/anthology/W17-1601.pdf) for using identity categories as a variables, particularly gender.
Describe the conditions under which the data was annotated (for example, if the annotators were crowdworkers, state what platform was used, or if the data was found, what website the data was found on). If compensation was provided, include that information here.
### Personal and Sensitive Information
State whether the dataset uses identity categories and, if so, how the information is used. Describe where this information comes from (i.e. self-reporting, collecting from profiles, inferring, etc.). See [Larson 2017](https://www.aclweb.org/anthology/W17-1601.pdf) for using identity categories as a variables, particularly gender. State whether the data is linked to individuals and whether those individuals can be identified in the dataset, either directly or indirectly (i.e., in combination with other data).
State whether the dataset contains other data that might be considered sensitive (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).
If efforts were made to anonymize the data, describe the anonymization process.
## Considerations for Using the Data
### Social Impact of Dataset
Please discuss some of the ways you believe the use of this dataset will impact society.
The statement should include both positive outlooks, such as outlining how technologies developed through its use may improve people's lives, and discuss the accompanying risks. These risks may range from making important decisions more opaque to people who are affected by the technology, to reinforcing existing harmful biases (whose specifics should be discussed in the next section), among other considerations.
Also describe in this section if the proposed dataset contains a low-resource or under-represented language. If this is the case or if this task has any impact on underserved communities, please elaborate here.
### Discussion of Biases
Provide descriptions of specific biases that are likely to be reflected in the data, and state whether any steps were taken to reduce their impact.
For Wikipedia text, see for example [Dinan et al 2020 on biases in Wikipedia (esp. Table 1)](https://arxiv.org/abs/2005.00614), or [Blodgett et al 2020](https://www.aclweb.org/anthology/2020.acl-main.485/) for a more general discussion of the topic.
If analyses have been run quantifying these biases, please add brief summaries and links to the studies here.
### Other Known Limitations
If studies of the datasets have outlined other limitations of the dataset, such as annotation artifacts, please outline and cite them here.
## Additional Information
### Dataset Curators
List the people involved in collecting the dataset and their affiliation(s). If funding information is known, include it here.
### Licensing Information
Provide the license and link to the license webpage if available.
### Citation Information
Provide the [BibTex](http://www.bibtex.org/)-formatted reference for the dataset. For example:
```
@article{article_id,
author = {Author List},
title = {Dataset Paper Title},
journal = {Publication Venue},
year = {2525}
}
```
If the dataset has a [DOI](https://www.doi.org/), please provide it here.
### Contributions
[@cylee] added this dataset as part of the Hugging Face Dataset library tutorial (https://huggingface.co/course/chapter5/5?fw=tf). |
dalle-mini | null | @article{thomee2016yfcc100m,
author = "Bart Thomee and David A. Shamma and Gerald Friedland and Benjamin Elizalde and Karl Ni and Douglas Poland and Damian Borth and Li-Jia Li",
title = "{YFCC100M}: The New Data in Multimedia Research",
journal = "Communications of the {ACM}",
volume = "59",
number = "2",
pages = "64--73",
year = "2016",
url = "http://cacm.acm.org/magazines/2016/2/197425-yfcc100m/fulltext",
} | The YFCC100M is one of the largest publicly and freely useable multimedia collection, containing the metadata of around 99.2 million photos and 0.8 million videos from Flickr, all of which were shared under one of the various Creative Commons licenses.
This version is a subset defined in openai/CLIP. | false | 475 | false | dalle-mini/YFCC100M_OpenAI_subset | 2021-08-26T17:56:01.000Z | null | false | 986e65392adb1f3bdab07c25ed9a23cb83a0b354 | [] | [
"arxiv:1503.01817"
] | https://huggingface.co/datasets/dalle-mini/YFCC100M_OpenAI_subset/resolve/main/README.md | # YFCC100M subset from OpenAI
Subset of [YFCC100M](https://arxiv.org/abs/1503.01817) used by OpenAI for [CLIP](https://github.com/openai/CLIP/blob/main/data/yfcc100m.md), filtered to contain only the images that we could retrieve.
| Split | train | validation |
| --- | --- | --- |
| Number of samples | 14,808,859 | 16,374 |
| Size | 1.9 TB | 2.1 GB |
Features:
* from the original dataset: `title`, `description`, `photoid`, `uid`, `unickname`, `datetaken`, `dateuploaded`, `capturedevice`, `usertags`, `machinetags`, `longitude`, `latitude`, `accuracy`, `pageurl`, `downloadurl`, `licensename`, `licenseurl`, `serverid`, `farmid`, `secret`, `secretoriginal`, `ext`, `marker`, `key`
* `img`: image content, can be loaded with `PIL.Image.open(io.BytesIO(item['img']))`
* `title_clean` and `description_clean`: derived from `title` and `description` using `clean_text` function detailed below
```python
def clean_text(text):
# decode url
text = urllib.parse.unquote_plus(text)
# remove html tags
text = re.sub('<[^<]+?>', '', text)
# remove multiple spaces + "\r" + "\n" + "\t"
text = " ".join(text.split())
return text
``` |
damlab | null | null | null | false | 319 | false | damlab/HIV_FLT | 2022-02-08T20:58:56.000Z | null | false | a8e47c9a43d12564240e175708fe4e9424d275f0 | [] | [] | https://huggingface.co/datasets/damlab/HIV_FLT/resolve/main/README.md | # Dataset Description
## Dataset Summary
This dataset was derived from the Los Alamos National Laboratory HIV sequence (LANL) database.
It contains the most recent version (2016-Full-genome), composed of 1,609 high-quality full-length genomes.
The genes within these sequences were processed using the GeneCutter tool and translated into corresponding amino acid sequences using the BioPython library Seq.translate function.
Supported Tasks and Leaderboards: None
Languages: English
## Dataset Structure
### Data Instances
Each column represents the protein amino acid sequence of the HIV genome.
The ID field indicates the Genbank reference ID for future cross-referencing.
There are 1,609 full length HIV genomes.
Data Fields: ID, gag, pol, env, nef, tat, rev, proteome
Data Splits: None
## Dataset Creation
Curation Rationale: This dataset was curated to train a model (HIV-BERT) designed to predict a variety of sequence-dependent features regarding HIV.
Initial Data Collection and Normalization: Dataset was downloaded and curated on 12/21/2021.
## Considerations for Using the Data
Social Impact of Dataset: This dataset can be used to study sequence-dependent features of HIV, a virus that has claimed the lives of many individuals globally in the last few decades.
Discussion of Biases: This dataset was derived from the Los Alamos National Laboratory HIV sequence (LANL) database full genome database and contains a representative sample from each subtype and geographic region.
## Additional Information:
- Dataset Curators: Will Dampier
- Citation Information: TBA |
damlab | null | null | null | false | 319 | false | damlab/HIV_PI | 2022-03-09T19:48:01.000Z | null | false | f0bada3a186a6ab795d578088eaff9cae1ee7106 | [] | [
"license:mit"
] | https://huggingface.co/datasets/damlab/HIV_PI/resolve/main/README.md | ---
license: mit
---
# Dataset Description
## Dataset Summary
This dataset was derived from the Stanford HIV Genotype-Phenotype database and contains 1,733 HIV protease sequences. A
pproximately half of the sequences are resistant to at least one antiretroviral therapeutic (ART).
Supported Tasks and Leaderboards: None
Languages: English
## Dataset Structure
### Data Instances
Each column represents the protein amino acid sequence of the HIV protease protein. The ID field indicates the Genbank reference ID for future cross-referencing. There are 1,733 total protease sequences.
Data Fields: ID, sequence, fold, FPV, IDV, NFV, SQV
Data Splits: None
## Dataset Creation
Curation Rationale: This dataset was curated to train a model (HIV-BERT-PI) designed to predict whether an HIV protease sequence would result in resistance to certain antiretroviral (ART) drugs.
Initial Data Collection and Normalization: Dataset was downloaded and curated on 12/21/2021.
## Considerations for Using the Data
Social Impact of Dataset: Due to the tendency of HIV to mutate, drug resistance is a common issue when attempting to treat those infected with HIV.
Protease inhibitors are a class of drugs that HIV is known to develop resistance via mutations.
Thus, by providing a collection of protease sequences known to be resistant to one or more drugs, this dataset provides a significant collection of data that could be utilized to perform computational analysis of protease resistance mutations.
Discussion of Biases: Due to the sampling nature of this database, it is predominantly composed of subtype B sequences from North America and Europe with only minor contributions of Subtype C, A, and D.
Currently, there was no effort made to balance the performance across these classes.
As such, one should consider refinement with additional sequences to perform well on non-B sequences.
## Additional Information:
- Dataset Curators: Will Dampier
- Citation Information: TBA
|
damlab | null | null | null | false | 321 | false | damlab/HIV_V3_bodysite | 2022-02-08T21:12:25.000Z | null | false | 7c81ad7c34d35f0ea4cabc28c24dc79c299dd6b3 | [] | [
"license:mit"
] | https://huggingface.co/datasets/damlab/HIV_V3_bodysite/resolve/main/README.md | # Dataset Description
## Dataset Summary
This dataset was derived from the Los Alamos National Laboratory HIV sequence (LANL) database.
It contains 5,510 unique V3 sequences, each annotated with its corresponding bodysite that it was associated with.
Supported Tasks and Leaderboards: None
Languages: English
## Dataset Structure
### Data Instances
Data Instances: Each column represents the protein amino acid sequence of the HIV V3 loop.
The ID field indicates the Genbank reference ID for future cross-referencing.
There are 2,935 total V3 sequences, with 91% being CCR5 tropic and 23% CXCR4 tropic.
Data Fields: ID, sequence, fold, periphery-tcell, periphery-monocyte, CNS, lung, breast-milk, gastric, male-genitals, female-genitals, umbilical-cord, organ
Data Splits: None
## Dataset Creation
Curation Rationale:
Initial Data Collection and Normalization: Dataset was downloaded and curated on 12/20/2021.
## Considerations for Using the Data
Social Impact of Dataset: This dataset can be used to study the mechanism by which HIV V3 loops allow for study of HIV compartmentalization.
Discussion of Biases: DDue to the sampling nature of this database, it is predominantly composed of subtype B sequences from North America and Europe with only minor contributions of Subtype C, A, and D.
Currently, there was no effort made to balance the performance across these classes.
As such, one should consider refinement with additional sequences to perform well on non-B sequences.
Additionally, this dataset is highly biased to peripheral T-cells.
## Additional Information:
- Dataset Curators: Will Dampier
- Citation Information: TBA
---
license: mit
--- |
damlab | null | null | null | false | 321 | false | damlab/HIV_V3_coreceptor | 2022-02-08T21:09:21.000Z | null | false | e6aae6b448d287929238c39a8bb880ae93ab4211 | [] | [] | https://huggingface.co/datasets/damlab/HIV_V3_coreceptor/resolve/main/README.md | # Dataset Description
## Dataset Summary
This dataset was derived from the Los Alamos National Laboratory HIV sequence (LANL) database.
It contains 2,935 HIV V3 loop protein sequences, which can interact with either CCR5 receptors on T-Cells or CXCR4 receptors on macrophages.
Supported Tasks and Leaderboards: None
Languages: English
## Dataset Structure
### Data Instances
Data Instances: Each column represents the protein amino acid sequence of the HIV V3 loop.
The ID field indicates the Genbank reference ID for future cross-referencing.
There are 2,935 total V3 sequences, with 91% being CCR5 tropic and 23% CXCR4 tropic.
Data Fields: ID, sequence, fold, CCR5, CXCR4
Data Splits: None
## Dataset Creation
Curation Rationale: This dataset was curated to train a model (HIV-BERT-V3) designed to predict whether an HIV V3 loop would be CCR5 or CXCR4 tropic.
Initial Data Collection and Normalization: Dataset was downloaded and curated on 12/20/2021.
## Considerations for Using the Data
Social Impact of Dataset: This dataset can be used to study the mechanism by which HIV V3 loops allow for entry into T-cells and macrophages.
Discussion of Biases: Due to the sampling nature of this database, it is predominantly composed of subtype B sequences from North America and Europe with only minor contributions of Subtype C, A, and D.
Currently, there was no effort made to balance the performance across these classes.
As such, one should consider refinement with additional sequences to perform well on non-B sequences.
## Additional Information:
- Dataset Curators: Will Dampier
- Citation Information: TBA
|
dansbecker | null | null | null | false | 320 | false | dansbecker/hackernews_hiring_posts | 2021-12-07T13:46:20.000Z | null | false | 68844f7ae036f6901f3b08526c45f6026ea26997 | [] | [] | https://huggingface.co/datasets/dansbecker/hackernews_hiring_posts/resolve/main/README.md | This dataset contains postings and comments from the following recurring threads on [Hacker News](http://news.ycombinator.com/)
1. Ask HN: Who is hiring?
2. Ask HN: Who wants to be hired?
3. Freelancer? Seeking freelancer?
These post types are stored in datasets called `hiring`, `wants_to_be_hired` and `freelancer` respectively.
Each type of posting has occurred on a regular basis for several years. You can identify when each comment/listing was added through the CommentTime field. The `ParentTitle` also indicates the date of the parent thread in text (e.g. `Ask HN: Who is hiring? (March 2021)`)
This dataset is not programmatically reproducible from source because it was uploaded as an experiment with HF datasets. The raw data was created by querying the public table `bigquery-public-data.hacker_news.full` in Google BigQuery.
Email addresses have been redacted from the dataset.
If this dataset is interesting/useful, I (Dan Becker) will look into improving reproducibility and other general clean-up.
This dataset may be useful for finding trends in tech and tech job listings. |
davanstrien | null | null | null | false | null | false | davanstrien/MOH | 2021-10-24T13:14:53.000Z | null | true | c87dbb18c233a2b410dd54de0bbf876a3bb0dcbb | [] | [] | https://huggingface.co/datasets/davanstrien/MOH/resolve/main/README.md | |
davanstrien | null | null | null | false | null | false | davanstrien/test | 2021-11-10T17:08:18.000Z | null | true | 8a1d1f51a14f5a5a5081e6165379ee48cd3c8cdc | [] | [
"annotations_creators:no-annotation",
"language_creators:other",
"licenses:cc0-1.0",
"multilinguality:multilingual",
"size_categories:unknown",
"source_datasets:original",
"task_categories:other",
"task_ids:language-modeling",
"task_ids:other-other-digital-humanities-research"
] | https://huggingface.co/datasets/davanstrien/test/resolve/main/README.md |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.