We explore the adoption of topic modeling to inform the seamless integration of community discovery and role analysis. For this purpose, we develop a new Bayesian probabilistic generative model of social media, according to which the observation of social links and textual contents is governed by novel and intuitive relationships among latent content topics, communities and roles. Variational inference under the devised model allows for exploratory, descriptive and predictive tasks, including the detection and interpretation of overlapping communities, roles and topics as well as the prediction of missing links. Extensive tests on real-world social media reveal a superior accuracy of the proposed model in comparison to state-of-the-art competitors, which substantiates the rationality of the motivating intuition. The experimental results are also insightfully inspected from a qualitative viewpoint.
展开▼
