NLP2SQL Using Semi-supervised Learning

Abstract

Human Computer interaction has been moving towards Natural language in the modern age. SQL (Structured Query Language) is the chief database query language used today. There are many flavors of SQL but all of them have the same basic underlying structure. This paper attempts to use the Natural Language inputs to query the databases, which is achieved by translating the natural language (which in our case is English) input into the SQL (specific to MySQL database) query language. Here we use a semi-supervised learning with Memory augmented policy optimization approach to solve this problem. This method uses the context of the natural language questions through database schema, and hence its not just generation of SQL code. We have used the WikiSQL dataset for all our experiments. The proposed method gives a 2.3% higher accuracy than the state of the art semi-supervised method on an average. © 2021, Springer Nature Singapore Pte Ltd.