Explicit Semantic Analysis for Enriching Content-Based User Profiles

A content-based recommender system suggests items similar to those previously liked by a user, therefore the recommendation process consists of matching up the features stored in a user profile with those of a content object (item). Usually a content-based user profile stores keywords that are more meaningful for that specific user. Common-sense knowledge could positively enrich that profile and content of items, thus helping to introduce more informative features than simple keywords. The idea of this work is to represent content objects, and consequentially user profiles, in terms of Wikipedia-based concepts.