CoSQL

该数据集是首个大规模、跨领域的对话式文本到SQL数据集，包含了近3000个对话，转化为超过30000个对话轮次以及10000个SQL查询。该数据集模拟了一个场景，其中标注者使用自然语言来提取数据库响应。其规模之大，涵盖了将近3000个对话、超过30000个轮次以及10000个查询任务，主要针对的是文本到SQL的任务。

This dataset is the first large-scale, cross-domain conversational text-to-SQL dataset. It contains nearly 3,000 dialogues, which are converted into over 30,000 dialogue turns and 10,000 SQL queries. The dataset simulates a scenario wherein annotators use natural language to extract responses from databases. With its substantial scale, it covers nearly 3,000 dialogues, more than 30,000 turns and 10,000 query tasks, and is primarily tailored for text-to-SQL tasks.