Spreading Activation Approach for Social Recommendations: The Case of Microblogging Services

Social media and social networks have received increasing attention by the scientific community due to their large amounts of user generated and enriched multimedia contents. They encourage interaction between users and their generated content, introduce new types of information for recommender systems. Meanwhile, social media bring heterogeneity to the networks. Traditionally entities of social networks are treated homogeneously, e.g. Messaging network, friendship network. However, real world social networks are intrinsically multidimensional and heterogeneous with different types of entities and different types of relationships. We suggest that different types of entities and relationships of social networks can be naturally and accurately modeled by heterogeneous networks, and propose a spreading activation approach for social recommendations. We employ the data of the microblogging service Twitter using our approach to recommend interesting people and topics to follow. The experimental results demonstrate the potential of our spreading activation approach.