earticle

영어

Recently microblog has become a popular Internet service. Yet, the abundance of micrblog floods users with huge volumes of information and poses a great challenge in terms of information overload. Recommender systems aim to alleviate information overload over microbog users by presenting the most attractive and relevant content. Many recommendation approaches have been proposed by leveraging the social relationships in microblog. To prove the feasibility of users’ social relationships as the bases of recommendation in microblog, we examine the correlations of social relationships and user interest similarities in microblog. Using real-world data set, we find that social relationship indicates positive connection with user interest similarity in microblog, but the positive connection is not strong. We also observe strong positive correlation between reciprocal social relationship and user interest similarity in microblog. As to users’ interaction, we find that the number of interaction between two users in social relation is a strong signal that controls users’ interest similarity. We then apply these findings to recommendation application in microblog to improve the accuracy of recommendation. The experiment on real-world data set shows that our findings are useful for recommendation.