HKUSTDial/NL2SQL_Handbook

Organizes Text-to-SQL research across a modular pipeline covering pre-processing, translation methods, and post-processing stages, with emphasis on LLM-based approaches. Provides a five-level challenge taxonomy and tracks evolution across four language model eras, from traditional NLP to modern large language models. Curates benchmarks, error analysis frameworks, and data synthesis techniques alongside peer-reviewed survey papers from TKDE, VLDB, and other top venues.

1,361 stars. Actively maintained with 2 commits in the last 30 days.