QUANTRA - Query understander and translator a natural language interface to relational database systems

A simple, yet powerful approach to providing a natural language interface to a database is presented. The developed system, called QUANTRA, answers queries given in English by translating them into SQL and executing the SQL queries against the database. It is easily configurable to any relational database with minimal manual intervention. Classical approaches to providing a natural language interface to databases include steps like syntax parsing, specifying logic predicates for keywords, generating an intermediate logical query and then obtaining SQL. These methods suffer from extensive knowledge coding phases, and in some cases are unable to handle complex queries. Recent approaches include graph theoretic techniques, where the query generation problem is viewed as a graph search problem. However, these methods suffer from ambiguity problems. They resort to statistical methods for disambiguation, which are not very reliable. QUANTRA strikes a balance between the two. It performs syntax parsing and uses graph theoretic methods. However, there is no extensive knowledge coding required and there is no use of statistical methods. Thus, it retains the advantages of both classical and recent methods while overcoming the limitations of the two.