arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2409.12172cs.CL
收藏

You Only Read Once (YORO):学习内化数据库知识以实现Text-to-SQL

You Only Read Once (YORO): Learning to Internalize Database Knowledge for Text-to-SQL

AWS AI(AWS人工智能)

Hideo Kobayashi, Wuwei Lan, Peng Shi, Shuaichen Chang, Jiang Guo, Henghui Zhu, Zhiguo Wang, Patrick Ng

AI总结:

针对Text-to-SQL任务中重复编码数据库模式导致推理成本高、易忽略关键知识的问题,提出YORO范式,训练时将数据库知识内化到模型参数中,推理无需模式编码,大幅缩短输入长度,在多基准上性能可比传统系统,大数据库及含缩写的数值检索问题上表现更优。

摘要:

尽管Text-to-SQL(文本转SQL)任务已取得显著进展,但现有解决方案会针对每个问题重复编码相同的数据库模式,导致不必要的高推理成本,且常常忽略关键的数据库知识。为解决这些问题,我们提出You Only Read Once(YORO,你只需读取一次)这一全新范式,它在训练阶段将数据库知识直接内化到Text-to-SQL模型的参数化知识中,推理阶段无需再进行模式编码。YORO将输入token长度大幅降低了66%-98%。尽管输入更短,我们的实验结果表明,YORO在三个基准上的性能与传统系统相当,且在大型数据库上表现显著更优。此外,YORO在处理涉及缩写等具有挑战性的数值检索的问题时表现出色。

英文摘要:

While significant progress has been made on the text-to-SQL task, recent solutions repeatedly encode the same database schema for every question, resulting in unnecessary high inference cost and often overlooking crucial database knowledge. To address these issues, we propose You Only Read Once (YORO), a novel paradigm that directly internalizes database knowledge into the parametric knowledge of a text-to-SQL model during training and eliminates the need for schema encoding during inference. YORO significantly reduces the input token length by 66%-98%. Despite its shorter inputs, our empirical results demonstrate YORO's competitive performances with traditional systems on three benchmarks as well as its significant outperformance on large databases. Furthermore, YORO excels in handling questions with challenging value retrievals such as abbreviation.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑