Towards context-based search engine selection

A well-known problem for web search is targeting search on information that satisfies users' information needs. User queries tend to be short, and hence often ambiguous, which can lead to inappropriate results from general-purpose search engines. This has led to a number of methods for narrowing queries by adding information. This paper presents an alternative approach that aims to improve query results by using knowledge of a user's current activities to select search engines relevant to their information needs, exploiting the proliferation of high-quality special-purpose search services. The paper introduces the PRISM source selection system and describes its approach. It then describes two initial experiments testing the system's methods.