Social network mashup: Ontology-based social network integration for statistic learning

The proliferation of online social websites results in the accumulation of a large volume of real-world data capturing social networks in diversified application domains. However, social networks are always separated with each other that causes the data isolated island phenomenon, which becomes impedance to implementing complex data analysis that requires comprehensive data stored in several social networks. In this paper, we present a social network mashup approach that uses the Semantic Web technology to integrate heterogeneous social networks that contain richer semantics. Secondly, we propose a statistic learning approach that learns a Probabilistic Semantic Model (PSM) from semantic structures of social networks. This framework can utilize these accumulated and integrated data without losing semantics. Lastly, our approach is evaluated by a real-life application that combines LinkedIn and DBLP to predict collaborative colleague relation.