SQL-Exchange: Transforming SQL Queries Across Domains

2025-08-09

AI Summary

This work addresses the challenges of SQL query migration across heterogeneous database schemas and the limited effectiveness of in-context learning in cross-database text-to-SQL tasks. To this end, we propose SQL-Exchange, the first framework to systematically explore and validate the feasibility and efficacy of structure-preserving cross-schema SQL query mapping. Our method leverages a structural alignment mechanism that jointly incorporates schema syntactic matching and semantic consistency constraints, enabling faithful preservation of the source query’s structural skeleton while adaptively rewriting domain-specific elements (e.g., table names, column names, value domains) to conform to the target schema. Extensive experiments across multiple LLM families (e.g., LLaMA, Qwen) and standard text-to-SQL benchmarks (Spider, BIRD) demonstrate that context examples generated via our mapping significantly improve zero-shot and few-shot text-to-SQL accuracy—achieving an average gain of 12.7%. Moreover, SQL-Exchange exhibits strong robustness on complex nested queries and cross-domain schema migrations.

Technology Category

• Data Mining & Knowledge Management: Intelligent Query Processing
• Machine Learning: Structured Learning
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

We introduce SQL-Exchange, a framework for mapping SQL queries across different database schemas by preserving the source query structure while adapting domain-specific elements to align with the target schema. We investigate the conditions under which such mappings are feasible and beneficial, and examine their impact on enhancing the in-context learning performance of text-to-SQL systems as a downstream task. Our comprehensive evaluation across multiple model families and benchmark datasets--assessing structural alignment with source queries, execution validity on target databases, and semantic correctness--demonstrates that SQL-Exchange is effective across a wide range of schemas and query types. Our results further show that using mapped queries as in-context examples consistently improves text-to-SQL performance over using queries from the source schema.

Problem

• Mapping SQL queries across different database schemas
• Enhancing in-context learning for text-to-SQL systems
• Evaluating structural alignment and execution validity

Innovation

• Maps SQL queries across different database schemas
• Preserves source query structure, adapts domain elements
• Improves text-to-SQL performance with mapped queries