Using Clusters on the Vivisimo Web Search Engine

Web search engine interfaces typically return search results as lists of Web links accompanied by brief textual summaries. An improvement on this results interface design has been introduced with the use of clusters. Clusters represent groups of similar items and are used to aggregate search results. Operational Web interfaces containing cluster representations are relatively new. This paper presents research from a transaction log analysis of Vivisimo.com, which is a Web search engine that dynamically clusters the user’s search results. The primary research question is: what are the characteristics of cluster use among Vivisimo searchers? One week’s worth of data was analyzed from April 25 to May 2, 2004. The findings show that searchers frequently clicked on cluster labels although seldom manipulated the cluster tree. This research has implications for interface design utilizing clustering technology and Web retrieval.