The continuous growing and pervasive use of social media offers new and interesting research opportunities, including the analysis of social media users’ behaviour and interactions. Nowadays, interactions are not only limited to social relations, but also to reading and writing activities. Thus, multiple and complementary information sources are available for describing users. One task that could benefit from integrating those multiple sources is community detection. This paper aims at improving the quality of the detected communities by proposing a general methodology for integrating both heterogeneous information from social networks, and the semantics of asymmetric interactions. Experimental evaluation confirmed the differentiated impact that diverse information sources have on community quality, and how to improve it by combining both social and content-based information.