David Lowe, Faculty of Engineering, University of Technology, Sydney, P.O. Box 123, Broadway, Sydney, 2007, Australia. email@example.com
World Wide Web, Search, Relevance, Navigation, Context
Search engines provide an important mechanism for dealing with the exploding volume of information on the Web. Numerous sophisticated algorithms have been developed to improve both the performance of search engines and the quality of the results achieved (as measured by both precision of the results and the extent of recall). However the lack of effective consideration of the context of indexed pages limits the relevance of results obtained.
Our work supports the hypothesis that the relevance of search results can be improved through the inclusion of contextual information. This contextual information can be obtained by an automated analysis of the local information space surrounding a given candidate page - with the local space defined by the navigational structures inherent in the Web. In particular, we treat the content of linked pages as providing a useful context for a source page, and which can consequently be analysed and used to improve searching precision. The research described in this paper involves the development of a prototype search engine that uses this technique. We describe current approaches to searching, and the theory behind our approach. We then look at a prototype search engine that implements these concepts, and the subsequent evaluation of the improvement in information retrieval. The significance of this work includes a greatly improved understanding of context-based searching and a substantially enhanced Web searching capability.
[ Full Paper ] [ Presentation ] [ Proceedings ] [ AusWeb2K Home Page ]