Uncovering user interests play an important role to develop personalized systems in various fields, including Web and pervasive computing. In particular, online social networks (OSNs) are being spotlighted as the means to understand users’ social behavior out of abundant online social information. Many researches have proposed methods for inferring users’ implicit interests in OSN based on social correlations between associated users and their social data. In this thesis, we propose a new inference method by taking into account the topics the user exchanged in terms of social activities with his/her social neighbors as well as their explicitly specified interests and familiarity with them. With 50 participant users in our experiment, a variety of ego-centric social network data for a year in the Facebook have been collected and analyzed. We demonstrate that our method outperforms the existing methods in inferring users’ interests. We also analyze Correlation-Weight and Interest-Weight, factors affecting the inference algorithm, to validate the effectiveness of them in a statistical way. For these tasks, we conduct a questionnaire survey that consists of three parts. We confirmed a distinct possibility to enhance the accuracy by leveraging multilateral communication contents and topic distributions extracted from them.