An Algorithm for Solving Natural Language Query Execution Problems on Relational Databases

Abstract

— There continues to be an increased need for non-experts interaction with databases. This is essential in their quest to make appropriate business decisions. Researchers have, over the years, continued to find a methodology that bridges the gap that exist between information need and users satisfaction. This has been the core in studies related to natural language information retrieval. In this paper, we understudy the existing methodology and develop a model to extend the proposition of (a) Bhardwaj et al where a MAPPER was developed and implemented on student database and (b) Nihalani et al. where an integrated interface was used on relational databases. We present a time saving executable algorithm that satisfies needed conditions required to retrieve results of natural language based queries from relational databases. Results of the experiment shows that the performance index of the algorithm is satisfactory and can be improved upon increasing the metadata table of the relational database. This is a sharp diversion from the keyword based search that has dominated most commercial databases in use today. The implementation was deployed in PHP and the retrieval time has compared favorably with earlier deployed models. We further propose the extension of this work in the areas of inculcating some fuzzy constraints to handle uncertainty and ambiguity which are inherent in human natural language.