Ontology based searches for Relational Databases

Abstract

Knowledge discovery applications and database driven applications require users to write structured queries to retrieve the necessary data from the relational database tables. Despite the fact that high level query languages are powerful and robust for this purpose, the user is expected to have a thorough knowledge of the query language and the underlying structure of the database in order to achieve his tasks. This complexity can be addressed if the users can search the content by means of normal text, which can be used to generate powerful semantic rich queries to generate the results that the user wants. The objective of this research project is to propose a mechanism that would facilitate plain text searches in relational databases. The proposed mechanism facilitates searches using plain text by making use of the concept of Ontologies, since Ontologies have the ability to better capture and structure Meta data. The proposed solution is a step by step process to transform a given relational database schema into a semantically rich Ontology. By transforming the databases to a form of Ontology, many benefits can be achieved. Some of the benefits are better responses for queries compared to traditional SQL, handling heterogeneity of databases and to provide abstraction to the user from the complex underlying structure of the database and the syntax and semantics of a high level query language. Further the Ontology created by transformation can be further enriched using other ontologies of that particular domain, enabling even better responses to queries. The proposed solution was evaluated by building a prototype and applying it on a sample Database schema. The prototype was able to successfully identify concepts from the database and relationships amongst them by extracting Meta data from the given relational database schema.