Information Retrieval from database queries

Abstract

Databases and documents are usually confined into separated environments inside organizations, controlled by Database Management Systems (DBMS) and Information Retrieval Systems (IRS), respectively. However, both DBMS and IRS frequently store data about the same entities, in this way presenting opportunities for integration. We propose a framework for DBMS-IRS integration that uses top ranked terms from a database query result as keywords for an IRS search, thus retrieving documents strongly related to the query. Indeed, the framework uses the ranked terms to “expand” an initial keyword search provided by the user. Moreover, our term ranking method measures the utility of a term through its dispersion along a query result, exploiting the fact that the query provides exact answers to the information need. Our experiments have confirmed the superiority of the approach to DBMS-IRS integration, as well as the effectiveness of our term ranking method.