Dynamic user modeling in a Web store shell

We describe a framework for the dynamic revision of user models in an adaptive Web store shell, which tailors the suggestion of goods, as well as the interaction style, to the characteristics and interests of the individual user. The behavior of the user is unobtrusively monitored, by segmenting it on the basis of the focus spaces explored in the browsing activity, and the actions performed by the user are summarized into abstract facts. These facts are used for revising the user model via the interpretation process performed by a Bayesian Network which relates user features to user behavior.