Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

CRED-SQL: Enhancing Real-world Large Scale Database Text-to-SQL Parsing through Cluster Retrieval and Execution Description

Created by
  • Haebom

Author

Shaoming Duan, Zirui Wang, Chuanyi Liu, Zhibin Zhu, Yuhao Zhang, Peiyi Han, Liang Yan, Zewu Penge

Outline

This paper proposes the CRED-SQL framework to improve the accuracy of Text-to-SQL systems, which convert Natural Language Queries (NLQs) into SQL queries in large-scale databases. Existing Text-to-SQL systems suffer from schema association and semantic drift issues due to semantically similar properties in large-scale databases, leading to reduced accuracy. CRED-SQL addresses these issues by accurately identifying tables and columns related to NLQs through cluster-based, large-scale schema search and introducing an intermediate representation language, Execution Description Language (EDL), between NLQ and SQL. This two-step process—translating NLQs into EDLs and EDLs into SQL—levers the powerful inference capabilities of LLMs while reducing semantic drift. Experimental results on two large-scale cross-domain benchmarks, SpiderUnion and BirdUnion, demonstrate that CRED-SQL achieves state-of-the-art performance.

Takeaways, Limitations

•
Takeaways:
◦
A novel method for improving the accuracy of text-to-SQL systems in large databases is presented.
◦
Effective schema matching and semantic drift reduction using cluster-based schema search and EDL.
◦
Leverage the powerful inference capabilities of LLMs to improve performance on Text-to-SQL tasks.
◦
Achieve state-of-the-art performance in SpiderUnion and BirdUnion benchmarks.
◦
Ensuring reproducibility and expandability of research through open source code disclosure.
•
Limitations:
◦
Further research is needed to determine the generalizability of the proposed framework.
◦
Need for performance evaluation and improvement for specific types of complex NLQs
◦
Research is needed to optimize EDL design and adapt it to various database schemas.
◦
Further experiments are needed to determine performance and scalability in real-world application environments.
👍
Made with Slashpage