Datasets:
The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
fever/fever(複製)
データの内容
英語の事実検証のデータ。Wikipedia から抜き出した文を書き換えて主張(claim)を作り、元の文を知らない状態で検証して、Supported・Refuted・NotEnoughInfo のいずれかのラベルを付けたもの。1 行が 1 件の主張と 1 件の根拠文の組で、根拠の Wikipedia の記事名(evidence_wiki_url)と文番号(evidence_sentence_id)を持つ。v2.0 は 2019 年の共有タスクで参加者が作った敵対的な主張を含む。
データの概要
| 項目 | 値 |
|---|---|
| 行数 | 5,845,480(3 構成・8 分割の合計) |
| 列数 | 3〜7(構成による) |
| 容量 | 23.23 KB(ファイル 3 本) |
| ファイルの形式 | py 1 |
| 言語 | en |
| 公開 | public |
構成と分割
| 構成(config) | 分割(split)ごとの行数 | 列数 |
|---|---|---|
v1.0 |
train 311,431、labelled_dev 37,566、unlabelled_dev 19,998、unlabelled_test 19,998、paper_dev 18,999、paper_test 18,567 |
7 |
v2.0 |
validation 2,384 |
7 |
wiki_pages |
wikipedia_pages 5,416,537 |
3 |
列(構成 v1.0。構成ごとに列が違う)
| 列 | 型 |
|---|---|
id |
int32 |
label |
string |
claim |
string |
evidence_annotation_id |
int32 |
evidence_id |
int32 |
evidence_wiki_url |
string |
evidence_sentence_id |
int32 |
行数と列は元のカードの YAML(dataset_info)の記載。データセットビューアでは読めなかった。全データセットの一覧は HYUKI5/dataset-catalog にある。
複製について
fever/fever をリビジョン 2a74f2909caf2b8656343aeb8203e50bf84dcb56 の時点でそのまま複製したもの。中身は変えていない。判断モデル Laya(convaiinnovations/laya)の微調整に使う候補として、調査メモ(Laya 学習用データセット調査)に載せたデータを集めている。
出典
- 元のデータセット: https://huggingface.co/datasets/fever/fever
- 複製したリビジョン:
2a74f2909caf2b8656343aeb8203e50bf84dcb56(2024-01-18 11:03:38+00:00) - 元のライセンス表記:
['cc-by-sa-3.0', 'gpl-3.0']
論文
- 元のカードに論文へのリンクは見つからなかった
アノテーションの方法
収集の根拠
初期のデータ収集と正規化
アノテーションの過程
微調整に使うときの注意
- 調査メモの台帳での区分: B(継承条件・gated・モデル由来の条件・上流の規約などが付く)。区分は記載の文言で機械的に付けたもので、法的な判定ではない。学習に入れる前に元のカードとライセンスの原文で確かめる。
- Laya の入力は 2,048 token まで(ModernBERT-large / mmBERT)。長い本文は先頭から黙って切られるので、切れる行は事前に除くか、要約・切り出しをしてから入れる。
- 分割は記事・文書単位で行い、同じ本文から作った別の問いを学習側と較正側に分けない。温度と閾値は学習に使っていない分割で別々に当てる。
調査メモの記述
| メモ | 節 | ライセンス(記載) | 確認 | 注記 |
|---|---|---|---|---|
general_english_nlu.md |
2. NLI・言い換え・事実検証・スタンス(noul / 3 択 choice) (L126) | cc-by-sa-3.0 | 確認済 | 原文注記: - fever/fever: FEVER(Thorne et al., NAACL 2018)。v1.0: train 311,431 / labelled_dev 37,566 / paper_dev 18,999 / |
元のカード(日本語訳)
以下は元のデータセットのカードの本文を日本語に訳したもの(機械翻訳)。訳に疑問があるときは、末尾の原文を確かめる。
"fever" のデータセットカード
目次
- データセットの説明
- データセットの概要
- 対応タスクとリーダーボード
- 言語
- データセットの構造
- データのインスタンス
- データフィールド
- データ分割
- データセットの作成
- 収集の根拠
- ソースデータ
- アノテーション
- 個人情報と機微な情報
- データの利用に関する考慮事項
- データセットの社会的影響
- 偏りの議論
- その他の既知の制限
- 追加情報
- データセットのキュレーター
- ライセンス情報
- 引用情報
- 貢献
データセットの説明
- ホームページ: https://fever.ai/
- リポジトリ: More Information Needed
- 論文: More Information Needed
- 問い合わせ先: More Information Needed
データセットの概要
ウェブ上の数十億のページが、考えうるほぼすべての話題について情報を提供しているため、考えうるほぼすべての質問に答える事実を集める能力があるはずである。しかし、この情報のごく一部しか構造化されたソース(Wikidata、Freebase など)には含まれていない。そのため、自由形式のテキストを構造化された知識に変換する能力によって制約を受けている。ただし、近年の研究とメディアの報道で焦点となっているもう一つの問題がある。信頼できないソースから出る誤った情報である。
FEVER ワークショップは、検証可能な知識抽出の研究の場であり、この方向への進展を促すためのものである。
FEVER データセット: FEVER(Fact Extraction and VERification)は、Wikipedia から抽出した文を書き換えて生成し、その後、元となった文を知らない状態で検証された 185,445 件の主張からなる。主張は Supported、Refuted、NotEnoughInfo に分類される。最初の 2 つのクラスについては、アノテーターは判断に必要な根拠となる文も記録している。
FEVER 2.0 敵対的攻撃データセット: FEVER 2.0 データセットは、2019 年の共有タスクの Breaker フェーズで参加者が提出した 1174 件の主張からなる。参加者(Breaker)は、既存のシステムに分類の誤りを起こさせる敵対的な例を生成する課題を与えられた。Breaker は、3 つのクラス(Supported、Refuted、NotEnoughInfo)それぞれに同数のインスタンスを含む、最大 1000 件のインスタンスからなるデータセットを提出した。共有タスクの有効な項目として認められたのは、新規の主張(すなわち元の FEVER データセットに含まれない主張)のみである。提出物はその後、正しさ(文法的に正しいこと、適切にラベル付けされていること、FEVER のアノテーションガイドラインの要件を満たすこと)について手作業で評価された。
対応タスクとリーダーボード
このタスクは、テキストの主張をテキストのソースに照らして検証することである。
テキスト含意(TE)・自然言語推論と比べたときの主な違いは、それらのタスクでは各主張を検証する文章が与えられ、近年は通常それが 1 文からなる点である。一方、検証システムでは、根拠を形成するためにその文章を大量の文書集合から検索する。
言語
データセットは英語である。
データセットの構造
データのインスタンス
v1.0
- ダウンロードしたデータセットファイルのサイズ: 44.86 MB
- 生成されたデータセットのサイズ: 40.05 MB
- ディスクの総使用量: 84.89 MB
'train' の例は次のようになる。
'claim': 'Nikolaj Coster-Waldau worked with the Fox Broadcasting Company.',
'evidence_wiki_url': 'Nikolaj_Coster-Waldau',
'label': 'SUPPORTS',
'id': 75397,
'evidence_id': 104971,
'evidence_sentence_id': 7,
'evidence_annotation_id': 92206}
v2.0
- ダウンロードしたデータセットファイルのサイズ: 0.39 MB
- 生成されたデータセットのサイズ: 0.30 MB
- ディスクの総使用量: 0.70 MB
'validation' の例は次のようになる。
{'claim': "There is a convicted statutory rapist called Chinatown's writer.",
'evidence_wiki_url': '',
'label': 'NOT ENOUGH INFO',
'id': 500000,
'evidence_id': -1,
'evidence_sentence_id': -1,
'evidence_annotation_id': 269158}
wiki_pages
- ダウンロードしたデータセットファイルのサイズ: 1.71 GB
- 生成されたデータセットのサイズ: 7.25 GB
- ディスクの総使用量: 8.97 GB
'wikipedia_pages' の例は次のようになる。
{'text': 'The following are the football -LRB- soccer -RRB- events of the year 1928 throughout the world . ',
'lines': '0\tThe following are the football -LRB- soccer -RRB- events of the year 1928 throughout the world .\n1\t',
'id': '1928_in_association_football'}
データフィールド
すべての分割でデータフィールドは同じである。
v1.0
id:int32型の feature。label:string型の feature。claim:string型の feature。evidence_annotation_id:int32型の feature。evidence_id:int32型の feature。evidence_wiki_url:string型の feature。evidence_sentence_id:int32型の feature。
v2.0
id:int32型の feature。label:string型の feature。claim:string型の feature。evidence_annotation_id:int32型の feature。evidence_id:int32型の feature。evidence_wiki_url:string型の feature。evidence_sentence_id:int32型の feature。
wiki_pages
id:string型の feature。text:string型の feature。lines:string型の feature。
データ分割
v1.0
| train | unlabelled_dev | labelled_dev | paper_dev | unlabelled_test | paper_test | |
|---|---|---|---|---|---|---|
| v1.0 | 311431 | 19998 | 37566 | 18999 | 19998 | 18567 |
v2.0
| validation | |
|---|---|
| v2.0 | 2384 |
wiki_pages
| wikipedia_pages | |
|---|---|
| wiki_pages | 5416537 |
データセットの作成
収集の根拠
ソースデータ
初期のデータ収集と正規化
原言語の生成者は誰か?
アノテーション
アノテーションの過程
アノテーターは誰か?
個人情報と機微な情報
データの利用に関する考慮事項
データセットの社会的影響
偏りの議論
その他の既知の制限
追加情報
データセットのキュレーター
ライセンス情報
FEVER のライセンス:
These data annotations incorporate material from Wikipedia, which is licensed pursuant to the Wikipedia Copyright Policy. These annotations are made available under the license terms described on the applicable Wikipedia article pages, or, where Wikipedia license terms are unavailable, under the Creative Commons Attribution-ShareAlike License (version 3.0), available at http://creativecommons.org/licenses/by-sa/3.0/ (collectively, the “License Termsâ€). You may not use these files except in compliance with the applicable License Terms.
引用情報
「FEVER Dataset」を使用する場合は、次を引用すること:
@inproceedings{Thorne18Fever,
author = {Thorne, James and Vlachos, Andreas and Christodoulopoulos, Christos and Mittal, Arpit},
title = {{FEVER}: a Large-scale Dataset for Fact Extraction and {VERification}},
booktitle = {NAACL-HLT},
year = {2018}
}
「FEVER 2.0 Adversarial Attacks Dataset」を使用する場合は、次を引用すること:
@inproceedings{Thorne19FEVER2,
author = {Thorne, James and Vlachos, Andreas and Cocarascu, Oana and Christodoulopoulos, Christos and Mittal, Arpit},
title = {The {FEVER2.0} Shared Task},
booktitle = {Proceedings of the Second Workshop on {Fact Extraction and VERification (FEVER)}},
year = {2018}
}
貢献
このデータセットを追加してくれた @thomwolf、@lhoestq、@mariamabarham、@lewtun、@albertvillanova に感謝する。
原文(英語)
Dataset Card for "fever"
Dataset Summary
With billions of individual pages on the web providing information on almost every conceivable topic, we should have the ability to collect facts that answer almost every conceivable question. However, only a small fraction of this information is contained in structured sources (Wikidata, Freebase, etc.) – we are therefore limited by our ability to transform free-form text to structured knowledge. There is, however, another problem that has become the focus of a lot of recent research and media coverage: false information coming from unreliable sources.
The FEVER workshops are a venue for work in verifiable knowledge extraction and to stimulate progress in this direction.
FEVER Dataset: FEVER (Fact Extraction and VERification) consists of 185,445 claims generated by altering sentences extracted from Wikipedia and subsequently verified without knowledge of the sentence they were derived from. The claims are classified as Supported, Refuted or NotEnoughInfo. For the first two classes, the annotators also recorded the sentence(s) forming the necessary evidence for their judgment.
FEVER 2.0 Adversarial Attacks Dataset: The FEVER 2.0 Dataset consists of 1174 claims created by the submissions of participants in the Breaker phase of the 2019 shared task. Participants (Breakers) were tasked with generating adversarial examples that induce classification errors for the existing systems. Breakers submitted a dataset of up to 1000 instances with equal number of instances for each of the three classes (Supported, Refuted NotEnoughInfo). Only novel claims (i.e. not contained in the original FEVER dataset) were considered as valid entries to the shared task. The submissions were then manually evaluated for Correctness (grammatical, appropriately labeled and meet the FEVER annotation guidelines requirements).
Supported Tasks and Leaderboards
The task is verification of textual claims against textual sources.
When compared to textual entailment (TE)/natural language inference, the key difference is that in these tasks the passage to verify each claim is given, and in recent years it typically consists a single sentence, while in verification systems it is retrieved from a large set of documents in order to form the evidence.
Languages
The dataset is in English.
Dataset Structure
Data Instances
v1.0
- Size of downloaded dataset files: 44.86 MB
- Size of the generated dataset: 40.05 MB
- Total amount of disk used: 84.89 MB
An example of 'train' looks as follows.
'claim': 'Nikolaj Coster-Waldau worked with the Fox Broadcasting Company.',
'evidence_wiki_url': 'Nikolaj_Coster-Waldau',
'label': 'SUPPORTS',
'id': 75397,
'evidence_id': 104971,
'evidence_sentence_id': 7,
'evidence_annotation_id': 92206}
v2.0
- Size of downloaded dataset files: 0.39 MB
- Size of the generated dataset: 0.30 MB
- Total amount of disk used: 0.70 MB
An example of 'validation' looks as follows.
{'claim': "There is a convicted statutory rapist called Chinatown's writer.",
'evidence_wiki_url': '',
'label': 'NOT ENOUGH INFO',
'id': 500000,
'evidence_id': -1,
'evidence_sentence_id': -1,
'evidence_annotation_id': 269158}
wiki_pages
- Size of downloaded dataset files: 1.71 GB
- Size of the generated dataset: 7.25 GB
- Total amount of disk used: 8.97 GB
An example of 'wikipedia_pages' looks as follows.
{'text': 'The following are the football -LRB- soccer -RRB- events of the year 1928 throughout the world . ',
'lines': '0\tThe following are the football -LRB- soccer -RRB- events of the year 1928 throughout the world .\n1\t',
'id': '1928_in_association_football'}
Data Fields
The data fields are the same among all splits.
v1.0
id: aint32feature.label: astringfeature.claim: astringfeature.evidence_annotation_id: aint32feature.evidence_id: aint32feature.evidence_wiki_url: astringfeature.evidence_sentence_id: aint32feature.
v2.0
id: aint32feature.label: astringfeature.claim: astringfeature.evidence_annotation_id: aint32feature.evidence_id: aint32feature.evidence_wiki_url: astringfeature.evidence_sentence_id: aint32feature.
wiki_pages
id: astringfeature.text: astringfeature.lines: astringfeature.
Data Splits
v1.0
| train | unlabelled_dev | labelled_dev | paper_dev | unlabelled_test | paper_test | |
|---|---|---|---|---|---|---|
| v1.0 | 311431 | 19998 | 37566 | 18999 | 19998 | 18567 |
v2.0
| validation | |
|---|---|
| v2.0 | 2384 |
wiki_pages
| wikipedia_pages | |
|---|---|
| wiki_pages | 5416537 |
Dataset Creation
Curation Rationale
Source Data
Initial Data Collection and Normalization
Who are the source language producers?
Annotations
Annotation process
Who are the annotators?
Personal and Sensitive Information
Considerations for Using the Data
Social Impact of Dataset
Discussion of Biases
Other Known Limitations
Additional Information
Dataset Curators
Licensing Information
FEVER license:
These data annotations incorporate material from Wikipedia, which is licensed pursuant to the Wikipedia Copyright Policy. These annotations are made available under the license terms described on the applicable Wikipedia article pages, or, where Wikipedia license terms are unavailable, under the Creative Commons Attribution-ShareAlike License (version 3.0), available at http://creativecommons.org/licenses/by-sa/3.0/ (collectively, the “License Termsâ€). You may not use these files except in compliance with the applicable License Terms.
Citation Information
If you use "FEVER Dataset", please cite:
@inproceedings{Thorne18Fever,
author = {Thorne, James and Vlachos, Andreas and Christodoulopoulos, Christos and Mittal, Arpit},
title = {{FEVER}: a Large-scale Dataset for Fact Extraction and {VERification}},
booktitle = {NAACL-HLT},
year = {2018}
}
If you use "FEVER 2.0 Adversarial Attacks Dataset", please cite:
@inproceedings{Thorne19FEVER2,
author = {Thorne, James and Vlachos, Andreas and Cocarascu, Oana and Christodoulopoulos, Christos and Mittal, Arpit},
title = {The {FEVER2.0} Shared Task},
booktitle = {Proceedings of the Second Workshop on {Fact Extraction and VERification (FEVER)}},
year = {2018}
}
Contributions
Thanks to @thomwolf, @lhoestq, @mariamabarham, @lewtun, @albertvillanova for adding this dataset.
- Downloads last month
- 77