Abstract. This paper describes an approach to product recommendation that combines in a novel way content- and collaborative-based filtering techniques. The system helps the user to specify a query that filters out unwanted products in electronic catalogues (content-based). Moreover, if the query produces too many or no results, the system suggests useful query changes that save the gist of the original request. This process goes on iteratively till a reasonable number of products is selected. Then, the selected products are ranked exploiting a case base of recommendation sessions (collaborative-based). Among the user selected items the system ranks higher items that are similar to those selected by other users in similar sessions (twofold similarity). The approach has been applied to a web travel application and it has been evaluated with real users. The proposed approach: a) reduces dramatically the number of user queries, b) reduces the number of browsed products and c) the selected items are found first on the ranked list. 1

Abstract. This paper looks at feature selection for ordinal text classification. Typical applications are sentiment and opinion classification, where classes have relationships based on an ordinal scale. We show that standard feature selection using Information Gain (IG) fails to identify discriminatory features, particularly when they are distributed over multiple ordinal classes. This is because inter-class similarity, implicit in the ordinal scale, is not exploited during feature selection. The Probability Re-distribution Procedure (PRP), introduced in this paper, explicates inter-class similarity by revising feature distributions. It aims to influence feature selection by improving the ranking of features that are distributed over similar classes, relative to those distributed over dissimilar classes. Evaluations on three datasets illustrate that the PRP helps select features that result in significant improvements on classifier performance. Future work will focus on automated acquisition of inter-class similarity knowledge, with the aim of generalising the PRP for a wider class of problems. 1