Datasets:
Tasks:
Text Retrieval
Sub-tasks:
entity-linking-retrieval
Languages:
Chinese
Size:
1M - 10M
ArXiv:
License:
add data format + change into the template
Browse files
README.md
CHANGED
|
@@ -23,33 +23,74 @@ task_categories:
|
|
| 23 |
- text-retrieval
|
| 24 |
task_ids:
|
| 25 |
- entity-linking-retrieval
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
---
|
| 27 |
|
| 28 |
|
| 29 |
-
#
|
| 30 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
|
| 32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
|
| 34 |
Hansel is a high-quality human-annotated Chinese entity linking (EL) dataset, focusing on tail entities and emerging entities:
|
| 35 |
- The test set contains Few-shot (FS) and zero-shot (ZS) slices, has 10K examples and uses Wikidata as the corresponding knowledge base.
|
| 36 |
- The training and validation sets are from Wikipedia hyperlinks, useful for large-scale pretraining of Chinese EL systems.
|
| 37 |
|
| 38 |
-
Please see our [WSDM 2023](https://www.wsdm-conference.org/2023/) paper [**Hansel**](https://arxiv.org/abs/2207.13005) to learn more about our dataset.
|
| 39 |
-
|
| 40 |
-
## Dataset
|
| 41 |
|
| 42 |
-
##
|
| 43 |
-
|
| 44 |
-
| | # Mentions | # Entities | Domain |
|
| 45 |
-
| ---- | ---- | ---- | ---- |
|
| 46 |
-
| Train | 9,879,813 | 541,058 | Wikipedia |
|
| 47 |
-
| Validation | 9,674 | 6,320 | Wikipedia |
|
| 48 |
-
| Hansel-FS | 5,260 | 2,720 | News, Social Media |
|
| 49 |
-
| Hansel-ZS | 4,715 | 4,046 | News, Social Media, E-books, etc.|
|
| 50 |
|
| 51 |
|
| 52 |
-
### Data
|
| 53 |
|
| 54 |
{"id": "hansel-eval-zs-1463",
|
| 55 |
"text": "1905电影网讯 已经筹备了十余年的吉尔莫·德尔·托罗的《匹诺曹》，在上个月顺利被网飞公司买下，成为了流媒体巨头旗下的新片。近日，这部备受关注的影片确定了自己的档期：2021年。虽然具体时间未定，但影片却已经实实在在地向前迈出了一步。",
|
|
@@ -61,9 +102,18 @@ Please see our [WSDM 2023](https://www.wsdm-conference.org/2023/) paper [**Hanse
|
|
| 61 |
"domain": "news"
|
| 62 |
}
|
| 63 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 64 |
## Citation
|
| 65 |
|
| 66 |
-
If you
|
| 67 |
|
| 68 |
```bibtex
|
| 69 |
@misc{xu2022hansel,
|
|
|
|
| 23 |
- text-retrieval
|
| 24 |
task_ids:
|
| 25 |
- entity-linking-retrieval
|
| 26 |
+
dataset_info:
|
| 27 |
+
- config_name: wiki
|
| 28 |
+
features:
|
| 29 |
+
- name: id
|
| 30 |
+
dtype: string
|
| 31 |
+
- name: text
|
| 32 |
+
dtype: string
|
| 33 |
+
- name: start
|
| 34 |
+
dtype: int64
|
| 35 |
+
- name: end
|
| 36 |
+
dtype: int64
|
| 37 |
+
- name: mention
|
| 38 |
+
dtype: string
|
| 39 |
+
- name: gold_id
|
| 40 |
+
dtype: string
|
| 41 |
+
splits:
|
| 42 |
+
- name: train
|
| 43 |
+
- name: validation
|
| 44 |
+
- config_name: hansel
|
| 45 |
+
features:
|
| 46 |
+
- name: id
|
| 47 |
+
dtype: string
|
| 48 |
+
- name: text
|
| 49 |
+
dtype: string
|
| 50 |
+
- name: start
|
| 51 |
+
dtype: int64
|
| 52 |
+
- name: end
|
| 53 |
+
dtype: int64
|
| 54 |
+
- name: mention
|
| 55 |
+
dtype: string
|
| 56 |
+
- name: gold_id
|
| 57 |
+
dtype: string
|
| 58 |
+
- name: source
|
| 59 |
+
dtype: string
|
| 60 |
+
- name: domain
|
| 61 |
+
dtype: string
|
| 62 |
+
splits:
|
| 63 |
+
- name: test-few-shot
|
| 64 |
+
- name: test-zero-shot
|
| 65 |
---
|
| 66 |
|
| 67 |
|
| 68 |
+
# Dataset Card for "Hansel"
|
| 69 |
|
| 70 |
+
## Table of Contents
|
| 71 |
+
- [Table of Contents](#table-of-contents)
|
| 72 |
+
- [Dataset Description](#dataset-description)
|
| 73 |
+
- [Dataset Structure](#dataset-structure)
|
| 74 |
+
- [Data Instances](#data-instances)
|
| 75 |
+
- [Data Splits](#data-splits)
|
| 76 |
+
- [Citation](#citation)
|
| 77 |
|
| 78 |
+
|
| 79 |
+
## Dataset Description
|
| 80 |
+
|
| 81 |
+
- **Homepage:** https://github.com/HITsz-TMG/Hansel
|
| 82 |
+
- **Paper:** https://arxiv.org/abs/2207.13005
|
| 83 |
|
| 84 |
Hansel is a high-quality human-annotated Chinese entity linking (EL) dataset, focusing on tail entities and emerging entities:
|
| 85 |
- The test set contains Few-shot (FS) and zero-shot (ZS) slices, has 10K examples and uses Wikidata as the corresponding knowledge base.
|
| 86 |
- The training and validation sets are from Wikipedia hyperlinks, useful for large-scale pretraining of Chinese EL systems.
|
| 87 |
|
| 88 |
+
Please see our [WSDM 2023](https://www.wsdm-conference.org/2023/) paper [**"Hansel: A Chinese Few-Shot and Zero-Shot Entity Linking Benchmark"**](https://arxiv.org/abs/2207.13005) to learn more about our dataset.
|
|
|
|
|
|
|
| 89 |
|
| 90 |
+
## Dataset Structure
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 91 |
|
| 92 |
|
| 93 |
+
### Data Instances
|
| 94 |
|
| 95 |
{"id": "hansel-eval-zs-1463",
|
| 96 |
"text": "1905电影网讯 已经筹备了十余年的吉尔莫·德尔·托罗的《匹诺曹》，在上个月顺利被网飞公司买下，成为了流媒体巨头旗下的新片。近日，这部备受关注的影片确定了自己的档期：2021年。虽然具体时间未定，但影片却已经实实在在地向前迈出了一步。",
|
|
|
|
| 102 |
"domain": "news"
|
| 103 |
}
|
| 104 |
|
| 105 |
+
### Data Splits
|
| 106 |
+
|
| 107 |
+
| | # Mentions | # Entities | Domain |
|
| 108 |
+
| ---- | ---- | ---- | ---- |
|
| 109 |
+
| Train | 9,879,813 | 541,058 | Wikipedia |
|
| 110 |
+
| Validation | 9,674 | 6,320 | Wikipedia |
|
| 111 |
+
| Hansel-FS | 5,260 | 2,720 | News, Social Media |
|
| 112 |
+
| Hansel-ZS | 4,715 | 4,046 | News, Social Media, E-books, etc.|
|
| 113 |
+
|
| 114 |
## Citation
|
| 115 |
|
| 116 |
+
If you find our dataset useful, please cite us.
|
| 117 |
|
| 118 |
```bibtex
|
| 119 |
@misc{xu2022hansel,
|