Personalized Hierarchical Organization of Web Documents

Abstract

The web space is flat in nature because web documents are not organized and have no relationship defined among them. Identifying related documents relevant to a user's interest is very difficult. Although most search engines define an indexing structure upon web documents, the indexing structure is primarily designed for performance purpose and may not reflect any semantic relationship among the documents. User profile approaches adopted in existing search engines require the users to specify their interests explicitly.
In this paper, we describe our work on establishing a hierarchical view to web documents for each individual user. The system defines hierarchical clusters for a collection of conceptually similar documents. Profiles are created to capture the features of each cluster. We show how the profiles may evolve to adapt to a user's change of interests by capturing the way in which each user organizes his/her web documents via common file management operations. We believe that such operations are more intuitive for a user. We discuss implementation issues and illustrate its feasibility via detail simulated studies. Preliminary results on our prototype system also verify the validity of our simulated results.