Trust-aware media recommendation in heterogeneous social networks

Social network sites, such as Facebook and Twitter, are gaining increasing popularity nowadays by providing a convenient platform for sharing and consuming information of all kinds. While the ever increasing information sources on the social network sites hold tremendous promise, how to select user interested information becomes nontrivial as users are easily overloaded by a vast amount candidate information sources. Furthermore, as the information providers are autonomous entities in an open social network environment, they may spread information that is unreliable or completely fake. Hence, technological advances are in demand to recommend information sources to social network users that both match their interests and come from reliable information sources. We develop a novel social media recommendation framework, referred to as GCCR, to tackle the above central challenges. GCCR is coined based on the key technologies that supports the proposed framework: Graph summarization, Content-based approach, Clustering, and Recommendation. A user-centric strategy is adopted that exploits the historical behavior of a set of seed users as evidence to assess the trustworthiness of different information providers. A two-phase process that employs graph summarization and content-based clustering is developed to partition users into different interest groups. The interest group information is then used for recommendation purpose. We perform extensive experiments on real-world social network data to assess the effectiveness of the proposed GCCR framework.