A proposed architecture for dynamically built NLIDB systems

Article type: Research Article

Authors: Sarhan, Amany

Affiliations: Fcaulty of Engineering, Computers and Automatic Control Department, Tanta University, Egypt. E-mail: amany_m_sarhan@tanta.edu.eg

Abstract: Natural Language Interface to Database (NLIDB) systems have provided an easy access to database system without the need for the user to use formal query languages, such as SQL. Database query languages can be difficult to the non-expert users and learning these formal queries takes a lot of time. In NLIDB, users can type a question or a sentence in their natural language (such as English or Arabic) then it will be converted through a special natural language interface interrupter into formal database query. A major problem that faces the NLIDB designers is the identification of the tables that contain the required information and the desired attributes in the query. Most of the existing systems use static templates for the queries, in which tables' names are embedded in the template. However, this requires large code that considers all possible query templates. The main objective of this paper is to introduce a dynamic approach to determine the tables name that the attributes involved in the query belong to by representing the database schema as a graph which is used to determine the tables' names. This will minimize the templates used to build the queries thus minimizing the code size and effort to track and build all possible queries. We also propose a dynamic component that helps in building any NLIDB called database management. This component is linked to all the NLIDB system's components and will automatically build the dictionaries and database schema graph. The proposed approach was applied to a case study called Mall Shopper's Guide (MSG) to investigate its effectiveness.

Keywords: Natural language, database, semantic graph, NLIDB, parser, semantic dictionary

DOI: 10.3233/JAD-2009-0174

Journal: International Journal of Knowledge-based and Intelligent Engineering Systems, vol. 13, no. 2, pp. 59-70, 2009