Cluster-Based Database Selection Techniques for Routing Bibliographic Queries

Given the large number of databases on the Internet, it is increasingly difficult for users to identify the databases relevant to their queries. Instead of broadcasting a given query to all databases, one would like to intelligently select only a small subset of databases for evaluating the query in order to reduce the amount of network and I/O overheads. This problem, also known as query routing, can be divided into three sub-problems known as database selection, query evaluation, and result merging. In this paper, we address the database selection problem for routing bibliographic queries. By clustering bibliographic records and summarizing their statistics, we are able to construct a knowledge base for each database and use it for database selection. We have proposed different database selection techniques based on different combinations of clustering algorithms and database ranking formulas. All these techniques have been experimented using carefully constructed bibliographic datab..