clips/beir-nl-dbpedia-entity

BEIR-NL是一个荷兰语版本的BEIR基准测试，涵盖了从生物医学、金融文本到一般网络内容的各种领域。数据集包含多个任务，如事实核查、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，分别用于存储文档、查询和查询文档相关性判断。数据集的创建目的是为了促进荷兰语信息检索模型的发展，通过自动翻译BEIR数据集来实现。然而，数据集存在一些局限性，如缺乏原生荷兰语资源、数据污染问题以及基准测试的有效性随时间变化的风险。

BEIR-NL is a Dutch-translated version of the BEIR benchmark, covering various tasks and domains including fact-checking, question-answering, bio-medical IR, argument retrieval, duplicate question retrieval, citation-prediction, and entity retrieval. The dataset includes corpus, queries, and qrels files, all in JSONL format. BEIR-NL is designed for zero-shot evaluation of information retrieval models in Dutch, addressing the need for benchmarks in underrepresented languages. The creation of the dataset includes the translation process and discusses known limitations, such as the reliance on translated data rather than native Dutch resources. Licensing information and citation details are also provided.