biu-nlp/WEC-Eng

# WEC-Eng
A large-scale dataset for cross-document event coreference extracted from English Wikipedia. </br>

- **Repository (Code for generating WEC):** https://github.com/AlonEirew/extract-wec
- **Paper:** https://aclanthology.org/2021.naacl-main.198/

### Languages

English

## Load Dataset
You can read in WEC-Eng files as follows (using the **huggingface_hub** library):

```json
from huggingface_hub import hf_hub_url, cached_download
import json
REPO_ID = "datasets/biu-nlp/WEC-Eng"
splits_files = ["Dev_Event_gold_mentions_validated.json",
"Test_Event_gold_mentions_validated.json",
"Train_Event_gold_mentions.json"]
wec_eng = list()
for split_file in splits_files:
wec_eng.append(json.load(open(cached_download(
hf_hub_url(REPO_ID, split_file)), "r")))
```

## Dataset Structure

### Data Splits
- **Final version of the English CD event coreference dataset**<br>
- Train - Train_Event_gold_mentions.json
- Dev - Dev_Event_gold_mentions_validated.json
- Test - Test_Event_gold_mentions_validated.json

| | Train | Valid | Test |
| ----- | ------ | ----- | ---- |
| Clusters | 7,042 | 233 | 322 |
| Event Mentions | 40,529 | 1250 | 1,893 |

- **The non (within clusters) controlled version of the dataset (lexical diversity)**<br>
- All (experimental) - All_Event_gold_mentions_unfiltered.json

### Data Instances

```json
{
"coref_chain": 2293469,
"coref_link": "Family Values Tour 1998",
"doc_id": "House of Pain",
"mention_context": [
"From",
"then",
"on",
",",
"the",
"members",
"continued",
"their"
],
"mention_head": "Tour",
"mention_head_lemma": "Tour",
"mention_head_pos": "PROPN",
"mention_id": "108172",
"mention_index": 1,
"mention_ner": "UNK",
"mention_type": 8,
"predicted_coref_chain": null,
"sent_id": 2,
"tokens_number": [
50,
51,
52,
53
],
"tokens_str": "Family Values Tour 1998",
"topic_id": -1
}
```

### Data Fields

|Field|Value Type|Value|
|---|:---:|---|
|coref_chain|Numeric|Coreference chain/cluster ID|
|coref_link|String|Coreference link wikipeida page/article title|
|doc_id|String|Mention page/article title|
|mention_context|List[String]|Tokenized mention paragraph (including mention)|
|mention_head|String|Mention span head token|
|mention_head_lemma|String|Mention span head token lemma|
|mention_head_pos|String|Mention span head token POS|
|mention_id|String|Mention id|
|mention_index|Numeric|Mention index in json file|
|mention_ner|String|Mention NER|
|tokens_number|List[Numeric]|Mentions tokens ids within the context|
|tokens_str|String|Mention span text|
|topic_id|Ignore|Ignore|
|mention_type|Ignore|Ignore|
|predicted_coref_chain|Ignore|Ignore|
|sent_id|Ignore|Ignore|

## Citation
```
@inproceedings{eirew-etal-2021-wec,
title = "{WEC}: Deriving a Large-scale Cross-document Event Coreference dataset from {W}ikipedia",
author = "Eirew, Alon and
Cattan, Arie and
Dagan, Ido",
booktitle = "Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
month = jun,
year = "2021",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.naacl-main.198",
doi = "10.18653/v1/2021.naacl-main.198",
pages = "2498--2510",
abstract = "Cross-document event coreference resolution is a foundational task for NLP applications involving multi-text processing. However, existing corpora for this task are scarce and relatively small, while annotating only modest-size clusters of documents belonging to the same topic. To complement these resources and enhance future research, we present Wikipedia Event Coreference (WEC), an efficient methodology for gathering a large-scale dataset for cross-document event coreference from Wikipedia, where coreference links are not restricted within predefined topics. We apply this methodology to the English Wikipedia and extract our large-scale WEC-Eng dataset. Notably, our dataset creation method is generic and can be applied with relatively little effort to other Wikipedia languages. To set baseline results, we develop an algorithm that adapts components of state-of-the-art models for within-document coreference resolution to the cross-document setting. Our model is suitably efficient and outperforms previously published state-of-the-art results for the task.",
}
```

## License
We provide the following data sets under a <a href="https://creativecommons.org/licenses/by-sa/3.0/deed.en_US">Creative Commons Attribution-ShareAlike 3.0 Unported License</a>. It is based on content extracted from Wikipedia that is licensed under the Creative Commons Attribution-ShareAlike 3.0 Unported License

## Contact
If you have any questions please create a Github issue at https://github.com/AlonEirew/extract-wec.

原始信息汇总

数据集概述

数据集名称

WEC-Eng

数据集描述

WEC-Eng是一个大规模的跨文档事件共指数据集，从英文维基百科中提取。

数据集语言

英语

数据集加载

使用huggingface_hub库加载数据集的示例代码如下：

json from huggingface_hub import hf_hub_url, cached_download import json REPO_ID = "datasets/biu-nlp/WEC-Eng" splits_files = ["Dev_Event_gold_mentions_validated.json", "Test_Event_gold_mentions_validated.json", "Train_Event_gold_mentions.json"] wec_eng = list() for split_file in splits_files: wec_eng.append(json.load(open(cached_download( hf_hub_url(REPO_ID, split_file)), "r")))

数据集结构

数据分割

• 最终版本的英文CD事件共指数据集
  • 训练集 - Train_Event_gold_mentions.json
  • 开发集 - Dev_Event_gold_mentions_validated.json
  • 测试集 - Test_Event_gold_mentions_validated.json

	训练	验证	测试
集群	7,042	233	322
事件提及	40,529	1,250	1,893

• 数据集的非控制版本（词汇多样性）
  • 全部（实验性） - All_Event_gold_mentions_unfiltered.json

数据实例

json { "coref_chain": 2293469, "coref_link": "Family Values Tour 1998", "doc_id": "House of Pain", "mention_context": [ "From", "then", "on", ",", "the", "members", "continued", "their" ], "mention_head": "Tour", "mention_head_lemma": "Tour", "mention_head_pos": "PROPN", "mention_id": "108172", "mention_index": 1, "mention_ner": "UNK", "mention_type": 8, "predicted_coref_chain": null, "sent_id": 2, "tokens_number": [ 50, 51, 52, 53 ], "tokens_str": "Family Values Tour 1998", "topic_id": -1 }

数据字段

字段	值类型	值描述
coref_chain	Numeric	共指链/集群ID
coref_link	String	共指链接维基百科页面/文章标题
doc_id	String	提及页面/文章标题
mention_context	List[String]	提及段落的标记化文本（包括提及）
mention_head	String	提及跨度头部标记
mention_head_lemma	String	提及跨度头部标记词元
mention_head_pos	String	提及跨度头部标记POS
mention_id	String	提及ID
mention_index	Numeric	提及在JSON文件中的索引
mention_ner	String	提及NER
tokens_number	List[Numeric]	提及上下文中的标记ID
tokens_str	String	提及跨度文本
topic_id	Ignore	忽略
mention_type	Ignore	忽略
predicted_coref_chain	Ignore	忽略
sent_id	Ignore	忽略

引用信息

@inproceedings{eirew-etal-2021-wec, title = "{WEC}: Deriving a Large-scale Cross-document Event Coreference dataset from {W}ikipedia", author = "Eirew, Alon and Cattan, Arie and Dagan, Ido", booktitle = "Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies", month = jun, year = "2021", address = "Online", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2021.naacl-main.198", doi = "10.18653/v1/2021.naacl-main.198", pages = "2498--2510", abstract = "Cross-document event coreference resolution is a foundational task for NLP applications involving multi-text processing. However, existing corpora for this task are scarce and relatively small, while annotating only modest-size clusters of documents belonging to the same topic. To complement these resources and enhance future research, we present Wikipedia Event Coreference (WEC), an efficient methodology for gathering a large-scale dataset for cross-document event coreference from Wikipedia, where coreference links are not restricted within predefined topics. We apply this methodology to the English Wikipedia and extract our large-scale WEC-Eng dataset. Notably, our dataset creation method is generic and can be applied with relatively little effort to other Wikipedia languages. To set baseline results, we develop an algorithm that adapts components of state-of-the-art models for within-document coreference resolution to the cross-document setting. Our model is suitably efficient and outperforms previously published state-of-the-art results for the task.", }

许可证

本数据集根据<a href="https://creativecommons.org/licenses/by-sa/3.0/deed.en_US">Creative Commons Attribution-ShareAlike 3.0 Unported License</a>提供。数据基于从维基百科提取的内容，该内容根据Creative Commons Attribution-ShareAlike 3.0 Unported License授权。

搜集汇总

背景与挑战

背景概述

WEC-Eng是一个从英文维基百科提取的大规模跨文档事件共指消解数据集，由巴伊兰大学NLP实验室创建，包含约16.8万行数据，支持训练、开发和测试划分，用于自然语言处理中的跨文档事件关联分析。