mehrzad-shahin/BuiltBench-retrieval

该数据集是一个针对建筑资产信息对齐任务的文本嵌入模型基准测试数据集，包含corpus、data和queries三个配置。corpus配置包含文档的标题和文本，data配置包含查询ID、文档ID和分数，queries配置包含查询文本。数据集适用于问题回答任务，数据语言为英文，规模在1K到10K之间。This dataset is a benchmarking dataset for text embedding models in aligning built asset information, containing three configurations: corpus, data, and queries. The corpus configuration includes document titles and text, the data configuration includes query-id, corpus-id, and score, and the queries configuration includes query text. The dataset is suitable for question-answering tasks, with data in English and a size range of 1K<n<10K.