clips/beir-nl-scifact

BEIR-NL是BEIR基准的荷兰语翻译版本，涵盖了多个领域的信息检索任务，包括事实核查、问答、生物医学信息检索等。数据集结构包括corpus、queries和qrels文件，分别用于存储文档、查询和查询文档相关性判断。数据集创建的目的是为了促进荷兰语信息检索模型的发展，尽管它是通过自动翻译生成的，可能存在一些局限性，如缺乏原生荷兰语资源、数据污染和基准有效性随时间减弱的问题。

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. The dataset includes tasks such as fact-checking, question-answering, bio-medical information retrieval, argument retrieval, duplicate question retrieval, citation-prediction, and entity retrieval. The dataset structure follows the original BEIR benchmark, containing a corpus, queries, and relevance judgments files. The creation of BEIR-NL aims to promote the development of Dutch information retrieval models by automatically translating publicly accessible BEIR datasets into Dutch. Although BEIR-NL provides a benchmark for evaluating Dutch IR models, it relies on data translated from English, which limits its full representation of the linguistic nuances and cultural context of the Dutch language.