RYANSQL: Recursively Applying Sketch-based Slot Fillings for Complex Text-to-SQL in Cross-Domain Databases

Text-to-SQL is the problem of converting a user question into an SQL query,\nwhen the question and database are given. In this paper, we present a neural\nnetwork approach called RYANSQL (Recursively Yielding Annotation Network for\nSQL) to solve complex Text-to-SQL tasks for cross-domain databases. State-ment\nPosition Code (SPC) is defined to trans-form a nested SQL query into a set of\nnon-nested SELECT statements; a sketch-based slot filling approach is proposed\nto synthesize each SELECT statement for its corresponding SPC. Additionally,\ntwo input manipulation methods are presented to improve generation performance\nfurther. RYANSQL achieved 58.2% accuracy on the challenging Spider benchmark,\nwhich is a 3.2%p improvement over previous state-of-the-art approaches. At the\ntime of writing, RYANSQL achieves the first position on the Spider leaderboard.\n