Enhancing Web Based Services by Coupling Document Classification with User Profile

Current Web search engines are not able to adapt their operations to the evolving needs, interests and preferences of the users. To cope with this weakness, we developed a system able to classify HTML (or, XML) documents into user prespecified categories of interests. The system processes the user current profile and a set of representative documents - for each category of interest, and produces a classification schema - presented as a set of representative category vectors. The classification schema is then utilized in order to classify new incoming Web documents to one (or, more) of the prespecified categories of interest and, furthermore, provides the basis to update user profile characteristics. We demonstrate effectiveness of the proposed approach via experimental results based on two indicative collections of documents