United_Join
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/jkkummerfeld/text2sql-data/
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为“United_Join”，是一个新构建的具有挑战性的数据集，它通过合并来自Spider-Dev的数据以及Finegan-Dollak等人（2018年）准备的另一个数据集中的带有JOIN操作符的SQL查询而形成。该数据集包含362个中等难度样本，317个困难难度样本，以及763个超困难难度样本，其中超过75%的样本属于困难及以上难度级别。该数据集规模为1,442个问题-SQL对，任务集中在处理带有JOIN操作的复杂SQL查询的文本到SQL转换。
The dataset named United_Join is a newly constructed challenging dataset, formed by merging data from Spider-Dev and SQL queries with JOIN operators sourced from another dataset prepared by Finegan-Dollak et al. (2018). It includes 362 medium-difficulty samples, 317 hard-difficulty samples, and 763 ultra-hard-difficulty samples, with over 75% of the total samples falling into the hard and above difficulty tiers. The dataset comprises a total of 1,442 question-SQL pairs, and its primary task focuses on text-to-SQL conversion for complex SQL queries involving JOIN operations.
提供机构：
Authors of the paper
