An Automated System for Web Portal Personalization

This paper proposes a system for personalization of web portals. A speci c implementation is discussed in reference to a web portal containing a news feed service. Techniques are proposed for e ective categorization, management, and personalization of news feeds obtained from a live news wire service. The process consists of two steps: rst manual input is required to build the domain knowledge which could be site-speci c; then the automated component uses this domain knowledge in order to perform the personalization, categorization and presentation. E ective schemes for advertising are proposed, where the targeting is done using both the information about the user and the content of the web page on which the advertising icon appears. Automated techniques for identifying sudden variations in news patterns are described; these may be used for supporting news-alerts. A description of a version of this software for our customer web site is provided.