Collaborative information filtering by using categorized bookmarks on the web

A bookmark means the URL information stored for memorizing a user's own footprints and revisiting that website. This paper regards this bookmark as one of the most meaningful information representing user preferences. An original bookmark indicating only address information is categorized for merging semantic meanings by using public web directory services. These categorized bookmarks are expressed in a hierarchical tree structure. However, most current web directory services cannot afford to normalize and manage the topic hierarchy. There are several kinds of structural incompleteness such as multiple references and heterogeneous tree structures. In order to extract user prefer-ences, this paper proposes a method for driving these problems and the influ-ence propagation methods based on Bayesian networks. Therefore, the preference maps representing users' interests are also established as tree structures. With respect to the user clustering, an approximate tree matching method is used for mapping (overlapping) users' preference maps. It is possible to make queries and process them efficiently according to categories. Finally, this paper is applied to implement collaborative web browsing that can guide and explore the web efficiently and adaptively.