Combining tag correlation and user social relation for microblog recommendation

With the development of social networking applications, microblog has turned to be an indispensable online communication network in our daily life. For microblog users, recommending high quality information is a demanding service. Some microblog services encourage users to annotate themselves with tags, which are used to describe their interests or attributes. However, few users are willing to create tags and available tags are not fully exploited for microblog recommendation. Besides, following/follower relationship in microblog is asymmetric, which can be used not only for communicating with friends or acquaintances but also for getting information on particular subjects. So far, there is no microblog recommendation algorithm which employs all the above mentioned information. This paper aims to investigate a joint framework to combine tag correlation and user social relation for microblog recommendation. Our approach identifies users' interests via their personal tags and social relations. More specifically, a user tag retrieval strategy is established to add tags for users without or with few tags, and the user-tag matrix is then built and user-tag weights are then obtained. In order to solve the problem of sparsity of the matrix, both inner and outer correlation between tags are investigated to update the user-tag matrix. Considering the significance of user social relation for microblog recommendation, a user user social relation similarity matrix is constructed. Moreover, an iterative updating scheme is developed to get the final tag-user matrix for computing the similarities between microblogs and users. We illustrate the capability of our algorithm by making experiments on real microblog datasets. Experimental results show that the algorithm is effective for microblog recommendation. (C) 2017 Elsevier Inc. All rights reserved.