Employing a domain ontology to gain insights into user behaviour ∗

As the Web becomes the de facto window shopping experience for the on-line customer so web personalization is becoming an integral part of many on-line retailers' customer relationship management (CRM) strategy. However Web personalization becomes increasingly difficult as the sheer size and heterogeneous nature of the information available on the web leads to information overload. In this paper we use an on-line movie retailer as a case study. We investigate how web visitor usage data may be combined with semantic domain knowledge to provide a deeper understanding of user behaviour. Our belief is that if we can explain the reasons for the users observed behaviour, we should be able to improve the quality of recommendations generated in the personalization process. In particular we introduce an " impact " measure, based on information theory. The measure captures the influence of a given concept from the domain ontology on user behaviour. We then combine impact measures for each of the concepts within the ontology to create an ontological profile for a user that can be used to personalize future interactions.