clips/beir-nl-fever

BEIR-NL是一个荷兰语版本的BEIR基准测试，涵盖了多个领域的信息检索任务，包括事实核查、问答、生物医学信息检索、论点检索、重复问题检索、引用预测和实体检索。数据集结构包括corpus、queries和qrels文件，分别包含文档、查询和查询文档相关性判断。数据集通过自动翻译BEIR数据集生成，但存在非原生荷兰语资源、数据污染和基准有效性随时间变化等局限性。

BEIR-NL is a Dutch-translated version of the BEIR benchmark, a diverse and heterogeneous collection of datasets covering various domains from biomedical and financial texts to general web content. BEIR-NL contains the following tasks: fact-checking (e.g., FEVER, Climate-FEVER, SciFact), question-answering (e.g., NQ, HotpotQA, FiQA-2018), bio-medical IR (e.g., TREC-COVID, NFCorpus), argument retrieval (e.g., Touche-2020, ArguAna), duplicate question retrieval (e.g., Quora, CqaDupstack), citation-prediction (e.g., SCIDOCS), and entity retrieval (e.g., DBPedia). The dataset adheres to the structure of the original BEIR benchmark, containing a corpus, queries, and qrels (relevance judgments file). BEIR-NL aims to provide zero-shot evaluation for Dutch information retrieval models, addressing the need for benchmarks in underrepresented languages like Dutch.