Using Hierarchical Clustering and Summarisation Approaches for Web Retrieval: Glasgow at the TREC 2002 Interactive Track.
Richard Osdin, Iadh Ounis, Ryen W. White · 2002
Introduction and Motivation Current search engines axe typified as having a lack of precision, coupled with an elongated ranked list style of result presentation. When combined, these factors make relevant data extraction increasingly complex. The main investigation of our participation in the Interactive Track of TREC 2002 is to assess the effectiveness of new visualisation techniques for displaying the results of search engines. Our current system, provisionally named HuddleSearch, uses a newly developed clustering algorithm, which dynamically organises the relevant documents into a traversable hierarchy of general to more-specific cluster categories. We have extended our TREC-10 summarisation tool to also allow the summarisation of multiple documents; whereby a summary paints a caricature of the contents of a cluster, rather than an individual document, thus allowing the user to provisionally judge a cluster's relevance prior to viewing its contents. The interaction between the us