Latent semantic social graph model for expert discovery in Facebook

Abstract

Expert finding systems employ social networks analysis and natural language processing to identify candidate experts in organization or enterprise datasets based on a user's profile, her documents, and her interaction with other users. Expert discovery in public social networks such as Facebook faces the challenges of matching users to a wide range of expertise areas, because of the diverse human interests. In this paper we analyze the social graph and the user's interactions in the form of posts and group memberships to model user interests and fields of expertise. The proposed model reflects expertise and interests of users based on experimental analysis of the explicit and implicit social data in Online Social Networks (OSNs). It employs social networks analysis, text mining, text classification, and semantic text similarity techniques to analyze and discover the latent semantic social graph model that can express user's expertise. The proposed model also considers the semantic similarity between user's posts and his groups, Influence of friendship on group's membership, and Influence of friendship on user's posts. Experiments on the Facebook data show significant validity of the proposed model.