Abstract

We present a new statistical approach, called principal clusters analysis, for analyzing millions of user navigations among Web documents. This technique can identify distinct clusters of related information on a given topic. In addition, it can determine which information items within a cluster are useful starting points to explore the topic of the cluster, as well as key documents within the cluster to explore the topic in greater detail. This technique should prove promising in addressing information overload and other knowledge management issues.

Share

COinS