The authors model schema knowledge as a graph mined from parsed query logs: tables connected by observed joins, with column-level domain entities extracted from query patterns.
Text-to-SQL works without schema documentation by mining query logs
Akgül et al. construct a knowledge graph from historical SQL queries and retrieve tables through domain entity matching, outperforming embedding-based retrieval in enterprise settings.
Big Tech
Ömer Faruk Akgül · Nikhil Kapahi · Prajit Reddy Muppidi
University of Southern California · Amazon
Research Digest··2 min read
Akgül, Kapahi and Muppidi present the first text-to-SQL framework that requires no schema documentation, learning database structure from historical query logs.
The Daily Edition
The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.