Building a Social Recommender System by Harvesting Social Relationships and Trust Scores between Users

Recommender systems were created to guide the user in a personalized way to interesting resources and to help users cope with the problem of information overload. A system's ability to adapt to the users' needs is based on gathering user-generated collective intelligence. In this paper, we present WSNRS, the system proposed for recommending content within social networks. The main goal of the system is to identify and filter the recently published valuable resources while taking into account the interactions and the relationships the user has within social structures. The interactions are logged and aggregated in order to determine the trust scores between users. Using the scores obtained, one can identify the types of relationship established between users; the scores will then be integrated into an adaptive global model used for recommending resources. Our approach presents several advantages over classic CF-based approaches and content-based recommendations regarding cold start, scalability and serendipitous recommendations. We will illustrate this with a case study that we made using data provided by the implementation of the system in a real online social network.