An ontology-based collaborative filtering recommendation for personalized TV program application

Even though recent hybrid methods have helped to avoid certain limitations of Content-based (CB) filtering and collaborative filtering (CF), scalability and sparsity are still major problems in large-scale recommendation systems. This paper presents a novel collaborative recommendation method based on ontology for personalized TV program application. In the proposed method, we give a new formula to calculate user ratings explicitly according to four TV-watching related behaviors' preferences. More specifically, in order to overcome sparsity problems of CF, we build the TV programs ontology to take semantic relationship to estimate contents' concept similarity. Experiments show that Ontology can explicitly reveal the similarity of programs. In our WEB-TV prototype system, Ontology-based collaborative recommendation method illustrate the ability improving the precision and recall in a large degree.