数据集:

BeIR/fever-qrels

英文

BEIR Benchmark数据集卡片

数据集概述

BEIR是一个异构基准数据集,由18个不同的数据集构成,代表了9个信息检索任务:

所有这些数据集都经过预处理,可用于实验。

支持的任务和排行榜

该数据集支持排行榜,评估模型的任务特定指标(如F1或EM),以及模型从维基百科中检索支持性信息的能力。

目前最佳性能的模型可以在 here 找到。

语言

所有任务都是用英语(en)进行的。

数据集结构

所有BEIR数据集都必须包含语料库、查询和Qrels(相关性判断文件)。它们必须采用以下格式:

  • 语料库文件:一个.jsonl文件(jsonlines),包含一个字典列表,每个字典包含三个字段_id(唯一文档标识符)、title(可选的文档标题)和text(文档段落或内容)。例如:{"_id": "doc1", "title": "阿尔伯特·爱因斯坦", "text": "阿尔伯特·爱因斯坦是德国出生的..."}
  • 查询文件:一个.jsonl文件(jsonlines),包含一个字典列表,每个字典有两个字段_id(唯一查询标识符)和text(查询文本)。例如:{"_id": "q1", "text": "谁发现了质能等效公式?"}
  • qrels文件:一个.tsv文件(以制表符分隔),包含三列,即查询id、语料库id和分数,按照此顺序。保留第一行作为标题。例如:q1 doc1 1

数据实例

BEIR数据集的一个高级示例:

corpus = {
    "doc1" : {
        "title": "Albert Einstein", 
        "text": "Albert Einstein was a German-born theoretical physicist. who developed the theory of relativity, \
                 one of the two pillars of modern physics (alongside quantum mechanics). His work is also known for \
                 its influence on the philosophy of science. He is best known to the general public for his mass–energy \
                 equivalence formula E = mc2, which has been dubbed 'the world's most famous equation'. He received the 1921 \
                 Nobel Prize in Physics 'for his services to theoretical physics, and especially for his discovery of the law \
                 of the photoelectric effect', a pivotal step in the development of quantum theory."
        },
    "doc2" : {
        "title": "", # Keep title an empty string if not present
        "text": "Wheat beer is a top-fermented beer which is brewed with a large proportion of wheat relative to the amount of \
                 malted barley. The two main varieties are German Weißbier and Belgian witbier; other types include Lambic (made\
                 with wild yeast), Berliner Weisse (a cloudy, sour beer), and Gose (a sour, salty beer)."
    },
}

queries = {
    "q1" : "Who developed the mass-energy equivalence formula?",
    "q2" : "Which beer is brewed with a large proportion of wheat?"
}

qrels = {
    "q1" : {"doc1": 1},
    "q2" : {"doc2": 1},
}

数据字段

所有配置的示例都具有以下特征:

语料库

  • corpus:表示文档标题和内容的字典特征,由以下组成:
    • _id:表示唯一文档id的字符串特征
      • title:字符串特征,表示文档标题。
      • text:字符串特征,表示文档内容。

查询

  • queries:表示查询的字典特征,由以下组成:
    • _id:表示唯一查询id的字符串特征
    • text:字符串特征,表示查询文本。

Qrels

  • qrels:表示查询文档相关性判断的字典特征,由以下组成:
    • _id:表示查询id的字符串特征
      • _id:字符串特征,表示文档id。
      • score:int32特征,表示查询和文档之间的相关性判断。

数据拆分

Dataset Website BEIR-Name Type Queries Corpus Rel D/Q Down-load md5
MSMARCO 12321321 msmarco train dev test 6,980 8.84M 1.1 12322321 444067daf65d982533ea17ebd59501e4
TREC-COVID 12323321 trec-covid test 50 171K 493.5 12324321 ce62140cb23feb9becf6270d0d1fe6d1
NFCorpus 12325321 nfcorpus train dev test 323 3.6K 38.2 12326321 a89dba18a62ef92f7d323ec890a0d38d
BioASQ 12327321 bioasq train test 500 14.91M 8.05 No 12328321
NQ 12329321 nq train test 3,452 2.68M 1.2 12330321 d4d3d2e48787a744b6f6e691ff534307
HotpotQA 12331321 hotpotqa train dev test 7,405 5.23M 2.0 12332321 f412724f78b0d91183a0e86805e16114
FiQA-2018 12333321 fiqa train dev test 648 57K 2.6 12334321 17918ed23cd04fb15047f73e6c3bd9d9
Signal-1M(RT) 12335321 signal1m test 97 2.86M 19.6 No 12336321
TREC-NEWS 12337321 trec-news test 57 595K 19.6 No 12338321
ArguAna 12339321 arguana test 1,406 8.67K 1.0 12340321 8ad3e3c2a5867cdced806d6503f29b99
Touche-2020 12341321 webis-touche2020 test 49 382K 19.0 12342321 46f650ba5a527fc69e0a6521c5a23563
CQADupstack 12343321 cqadupstack test 13,145 457K 1.4 12344321 4e41456d7df8ee7760a7f866133bda78
Quora 12345321 quora dev test 10,000 523K 1.6 12346321 18fb154900ba42a600f84b839c173167
DBPedia 12347321 dbpedia-entity dev test 400 4.63M 38.2 12348321 c2a39eb420a3164af735795df012ac2c
SCIDOCS 12349321 scidocs test 1,000 25K 4.9 12350321 38121350fc3a4d2f48850f6aff52e4a9
FEVER 12351321 fever train dev test 6,666 5.42M 1.2 12352321 5a818580227bfb4b35bb6fa46d9b6c03
Climate-FEVER 12353321 climate-fever test 1,535 5.42M 3.0 12354321 8b66f0a9126c521bae2bde127b4dc99d
SciFact 12355321 scifact train test 300 5K 1.1 12356321 5f7d1de60b170fc8027bb7898e2efca1
Robust04 12357321 robust04 test 249 528K 69.9 No 12358321

数据集创建

策划理由

[需要更多信息]

源数据

初始数据收集和规范化

[需要更多信息]

谁是源语言的生成者?

[需要更多信息]

注释

注释过程

[需要更多信息]

谁是注释者?

[需要更多信息]

个人和敏感信息

[需要更多信息]

使用数据的注意事项

数据集的社会影响

[需要更多信息]

偏见讨论

[需要更多信息]

其他已知限制

[需要更多信息]

其他信息

数据集策划者

[需要更多信息]

许可信息

[需要更多信息]

引文信息

引用方式:

@inproceedings{
thakur2021beir,
title={{BEIR}: A Heterogeneous Benchmark for Zero-shot Evaluation of Information Retrieval Models},
author={Nandan Thakur and Nils Reimers and Andreas R{\"u}ckl{\'e} and Abhishek Srivastava and Iryna Gurevych},
booktitle={Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2)},
year={2021},
url={https://openreview.net/forum?id=wCu6T5xFjeJ}
}

贡献

感谢 @Nthakur20 添加了该数据集。