Personalized Product Recommendation through Interactive Query Management and Case-Based Reasoning

This position paper describes a novel approach to the design of personalized recommender systems. The approach integrates content-based and collaborative filtering techniques, case-based reasoning, and an HCI perspective to system evaluation and user modeling. Following this method, we developed and tested a system prototype that helps the user to construct a travel plan by recommending promising items or by proposing entire plan templates. The system aids the user to specify a query that winnows out unwanted products in electronic catalogs and then sorts the results according to a case-based similarity metric. A case is a rich hierarchical data structure, containing a user’s profile information and all the items previously selected by that user. If the search fails, the system supports the refinement of the query, interacting with the user in a mixed-initiative approach. The system has been empirically evaluated in a pilot study and it is currently undergoing a tighter test in a new experimental effort. The results of the pilot evaluation are discussed, with special reference to the user-system interaction aspects