Bidirectional Attention for SQL Generation

Generating structural query language (SQL) queries from natural language is a long-standing open problem and has been attracting considerable interest recently, driven by the explosive development of deep learning techniques. Toward solving the problem, we leverage the structure of SQL queries and present a sketch-based approach or synthesizing way to solve this problem, which turns the solving method to a sequence-to-set problem and word order generation problem. We employ the bidirectional attention mechanisms and character level embedding to further improve the result. Experimental evaluations show that our model achieves the state-of-the-art results in WikiSQL dataset.