Markov Logic Network Based Social Relation Inference for Personalized Social Search

Most recommendation systems are based on historical data and profile files. The most common method is collaboration filtering. After analysis, a collaboration filtering recommender system differentiates one user profile from another to determine what to recommend. There has been substantial research on personalized social search. However, previous research has neglected semantic social information, making no use of definite relations between objects. This problem can be solved using ontology and inference rules. In this paper, Markov-logic-network (MLN)-based social relation inference is performed using social user information, such as country, age, and preference. In addition, this paper evaluates whether the inference results regarding social relations have been correctly predicted based on social user data. The user's personal and business relations are inferred based on MLNs and a social network comprised of user profile data.