ibm-research/dp-bench

DP-Bench（数据产品基准的简称）是首个用于评估数据产品创建系统的基准。它包含以下内容：- 特定业务用例的描述，称为数据产品请求（DPRs）；- 每个DPR对应的数据产品，包括相关数据库表和列的子集，以及从现有列派生的列；- 派生列的来源（SQL形式）；- 创建这些数据产品的实际数据库模式；- 与每个业务用例对应的自然语言问题；- DPRs和数据产品的标注主题。

DP-Bench (shortened from Data Product Benchmark) is the first of its kind benchmark for evaluating data product creation systems. It contains: - Description of specific business use cases, called data product requests (DPRs); - Corresponding data products for each DPR, consisting of relevant subsets of database tables and columns, as well as derived columns; - Provenance (in SQL) for the derived columns; - Actual DB schemas from which these data products were created; - Natural language questions corresponding to each business use case; - Annotated topics for the DPRs and data products.