Collaborative Web search is a community-based approach to adaptive Web search that is fundamentally case-based: the results of similar past search sessions are reused in response to new target queries. Previously, we have demonstrated that this approach to Web search can offer communities of like-minded searchers significant benefits when it comes to result relevance. In this paper we examine the fundamental issue of query similarity that drives the selection and reuse of previous search sessions. In the past we have proposed the use of a relatively simple form of query similarity, based on the overlap of query-terms. In this paper we examine and compare a collection of 10 alternative metrics that use different types of knowledge (query-terms vs. result-lists vs. selection behaviour) as the basis for similarity assessment.
展开▼
