News personalization using the CF-IDF semantic recommender

When recommending news items, most of the traditional algorithms are based on TF-IDF, i.e., a term-based weighting method which is mostly used in information retrieval and text mining. However, many new technologies have been made available since the introduction of TF-IDF. This paper proposes a new method for recommending news items based on TF-IDF and a domain ontology. It is demonstrated that adapting TF-IDF with the semantics of a domain ontology, resulting in Concept Frequency - Inverse Document Frequency (CF-IDF), yields better results than using the original TF-IDF method. CF-IDF is built and tested in Athena, a recommender extension to the Hermes news personalization framework. Athena employs a user pro??le to store concepts or terms found in news items browsed by the user. The framework recommends new articles to the user using a traditional TF-IDF recommender and the CFIDF recommender. A statistical evaluation of both methods shows that the use of an ontology signi??cantly improves the performance of a traditional recommender