Evaluating and enhancing meta-search performance in digital libraries

Applying meta search systems is a suitable method for supporting the user if there are many different retrieval services available on the Web. Due to information splitting strategies of literature services existing meta search systems either provide minimal integration of results or slow response times. We present an approach that combines techniques of personalization and query processing in order to satisfy the user's demand for both fast and comprehensive results. In order to evaluate and compare different query processing strategies and additional influencing parameters we developed a simulation tool called SIMPSON. Thereby, we can observe the performance of query processing within the context of different response times of the underlying digital library services in the Web, with different kinds of user queries, and with different sizes of query results. To evaluate and compare the performance of different query processing and duplicate detection strategies we developed metrics, particularly with regard to user satisfaction. We present results from our first experiments with SIMPSON, focusing on duplicate detection, query specification, and Web server performance of the underlying digital library services.