Designing a recommender system by using ontologies and semantic reasoning

Recommender systems research is becoming one of the popular research fields. Unlike search engines recommender systems are the systems which recommends the most suitable items to the users without any manual effort of the user. Recommender systems have really wide application areas like shopping, movies, music, book and news. Since every application area has its own challenges, it is really hard to develop a domain independent recommender system. There are many techniques for developing the recommender systems. Using the similarities between users or content or both is a common way to make recommendations. In this thesis, a recommender system using ontologies and semantic reasoning is proposed. The proposed system is developed in the news domain which has many additional challenges compared to other domains because of its dynamic nature. The existing news recommender systems do not take into account the news sources where the news articles come from. According to the research done by the media compaines, for a wider view of the news it is important to get the news articles from different sources. So in the news recommendation procedure it is important that the news sources are taken into account. In this thesis, a news source recommender system is proposed. Also the association rules are extracted by using the users' reading patterns. As a result, the relations between news sources gathered from this method are compared with the results from ontologies.