An intelligent natural language interface to relational databases

Abstract

This paper is a study on constructing a natural language interface to relational databases, which accepts natural language fuzzy questions as inputs and generates answers under the form of tables or short answers. Firstly, the question is parsed using a semantic grammar and then, it is translated into a SQL query using a set of translation rules. Finally, a database management system is left to find the result table with its own specialized optimization and planning techniques. Experimental results show that this approach can analyze a wide range of questions with high accuracy and produce reasonable answers.