Across the web : a collaborative approach to web page recommendation, classification, and automatic web directory building

Abstract

Two problems that web users sometimes have whilst navigating a web containing billions of documents are "lost in hyperspace" and "cognitive overload". Bookmarks are a way of tackling "lost in hyperspace" but tools are needed to help users manage their bookmarks. Web page recommender systems tackle "cognitive overload" by presenting the user with a number of pages that are highly relevant to the user's interest. AcrossTheWeb is a collaborative-based system that provides automatic bookmark classification and web page recommendation facilities. Topic segmentation techniques are used to identify page topics. Page segments are compared with user categories having matching topics. Similar categories are recommended as potentially good destinations for the incoming bookmark. Alternatively, when no similar categories are found, a new category is created and possible category names are proposed by analysing how other users have bookmarked similar pages. Web page recommendation is also topic based. AcrossTheWeb recommends pages from other users having similar interests. Comparison between users is done at the interest level, rather than at the user level. Thus, user categories are compared and the system recommends unseen pages from categories which are similar enough.