User Content Mining Supporting Usage Content for Web Personalization

In Web personalization usage mining has been used in combination of standard methods to help predict user needs based on their transaction histories. Although information in usage logs of a Web server reflects the interests of users to the site, the users are potentially not aware of all information needs that can be addressed through the Web server. User data (content) is a valuable source for discovering new users’ needs that can be satisfied by the server. This paper proposes a new framework that combines both usage content mining and user content mining for Web personalization. We use automatically extracted keyphrases as the main features for representing server usage data, server content data and publicly available user data. This uniform representation is used to support a finer granularity for the mining techniques that we use such as clustering and similarity computation.∗