A Metadata Search Approach to Keyword Query in Relational Databases

Abstract

This paper proposes an effective approach to keyword query in relational databases. It uses a semantic graph model consisting of database metadata, database values, user terms, and their semantic connections. Keywords of a query determine all possible connected subgraphs of the semantic model. A query answer is a subgraph with the minimum connections. In addition, the approach proposes to rank result tuples of the answer subgraph using the IR-style ranking function. Our experiment results show that queries with metadata terms give more precise answers than queries without them. General Terms