Nowadays, people use online social networks almost every day. They activate either due to their interests, or to search or catch their desirable information. Users of online social networks generate structural and contextual traces that can be analyzed by, i.e., network science researchers. Researchers can describe networks fabricated out of online traces from different perspectives that one of them is communities. Overlapping communities are overlapped structures, in which nodes have denser connections with each other than the rest of the network. Different approaches have addressed this problem; however, few analyses and methods have focused on contextual traces generated by users. As such, in this paper, we propose an algorithm that uses actual content produced by users. This algorithm uses term frequency of words generated by users and combines them by an extended clustering technique. Our evaluation results compare the proposed content-based community detection with structural-based methods. We also reveal community properties as well as its relation to contextual information. Administrators can use these algorithms in question & answer forums where the explicit links among users are missing.
展开▼
