Combining Linguistic Values and Semantics to Represent User Preferences

Since the advent of the Web 2.0, the amount of digital data available became increasingly overwhelming for a user looking for specific information. As a consequence, personalisation systems aiming at assisting the user in this task have emerged. The use of semantic web technologies to represent user profiles and their interests has shown some promising results allowing to infer preferences not directly gathered via explicit or implicit profiling. On the other hand, linguistic values, often conveniently used by humans when expressing their tastes or preferences, are another way to provide richer representation of user preferences. The aim of this paper is to propose a combination of semantic user modelling and linguistics values, and to show how a recommender system could benefit from this representation. To achieve this objective, we first propose an integrated semantic user model based on FOAF, permitting the expression of contextualised and weighted interests. An integration of linguistic values within this user model is then exemplified and, finally, we also propose an aggregation method to exploit linguistic values in recommender systems.