SOTAVerified

Datrics Text2SQL. A Framework for Natural Language to SQL Query Generation

2025-03-15- 2025Code Available2· sign in to hype

Tetiana Gladkykh, Kirykov Kyrylo

Code Available — Be the first to reproduce this paper.

Reproduce

Code

Abstract

Datrics Text2SQL is a Text-to-SQL framework using Retrieval-Augmented Generation (RAG) to enhance accuracy and reliability. By integrating domain knowledge, database structure, and example-based learning, it addresses common SQL generation challenges such as incorrect table selection, faulty joins, and misinterpreted business rules. A knowledge base built from database documentation and successful queries is stored in a vector database, enabling retrieval of relevant context for precise SQL generation. The structured process of knowledge collection, context retrieval, and query generation ensures adaptability to various database schemas and query types.

Tasks

Reproductions