Data Augmentation with Hierarchical SQL-to-Question Generation for Cross-domain Text-to-SQL Parsing

Data augmentation has attracted a lot of research attention in the deep learning era for its ability in alleviating data sparseness. The lack of data for unseen evaluation databases is exactly the major challenge for cross-domain text-to-SQL parsing. Previous works either require human intervention to guarantee the quality of generated data \cite{yu2018syntaxsqlnet}, or fail to handle complex SQL queries \cite{guo2018question}. This paper presents a simple yet effective data augmentation framework. First, given a database, we automatically produce a large amount of SQL queries based on an abstract syntax tree grammar \cite{yin2018tranx}. We require the generated queries cover at least 80\% of SQL patterns in the training data for better distribution matching. Second, we propose a hierarchical SQL-to-question generation model to obtain high-quality natural language questions, which is the major contribution of this work. Experiments on three cross-domain datasets, i.e., WikiSQL and Spider in English, and DuSQL in Chinese, show that our proposed data augmentation framework can consistently improve performance over strong baselines, and in particular the hierarchical generation model is the key for the improvement.