Identification of appropriate database relations through attributes extracted from natural language queries

Abstract

This paper presents a novel approach for recognizing database relations from a Natural Language Query (NLQ). Various significant phrases, which are extracted from NLQs for mapping to database details, have been used to isolate database relations with a goal to make available the actual information asked for by database users. A few novel algorithms have been incorporated to carry out this work. Moreover, application of common database operations like 'natural join' carried out when information comes from more than one database relation have been thoroughly investigated to be able to respond accurately to an NLQ. Correctness of the 'join' operation is determined by applying a number of algorithms which ensure the criteria of the operation even if the NLQ does not have enough information to isolate proper database relations. Besides, supervised learning techniques have been used to improve the process of recognizing relations. Some heavy duty database systems were used for experimental verification of the proposed methodology and the outcome of the experiments clearly shows that our proposed method can identify appropriate database relations through attributes extracted from NLQs to databases. © 2011 Springer-Verlag.