Summary

Clustering has been proposed as a comprehensive information access method. Search results clustering is about presenting the results of a search engine as a set (or a hierarchy) of thematic categories. Such presentation makes it possible to enhance the information given in the ranked lists which search engines produce. Typically users do not look at more than the top-ranking pages, and thus may easily miss some interesting pages buried deeper in the ranked list or some nontrivial relationships present in the results. We apply clustering to support users in interactive browsing through thematically organized search results. We present some clustering methods tailored to on-line processing of snippets, i.e. the short fragments of Web documents visible in the ranked list. Those methods take into account the time aspect, the particular requirements for clustering texts, and importantly, the readability of the produced grouping. Finally, we present the user interface of an actual system which automatically builds a navigable hierarchy of meaningful document groups. We believe that such system addresses two chief needs of the Web users: the need for efficient access to the up-to-date information on every available topic and the need for an organized and meaningful presentation of the desired information.