Semantic-based Followee Recommendations on Twitter Network

Twitter bloggers use the concept of follower/followee in order to inform and to be informed of all recent activities of users who have similar interests and preferences. Moreover, finding relevant users to follow becomes a crucial task due to the rapid growth of Twitter network and the huge number of daily registered users. Thus, the need for a system to assist users in such task is very important. Indeed, recent studies use lexical analysis to recommend people to follow. In this paper, we propose a followee recommender system based on semantic analysis of user profiles content by leveraging the follower/followee topology. We perform experiments using a real dataset harvested from Twitter. Experimental results show that our approach improves lexical-based approach by more than 5% on recall value for recommending 5 followees, proving that dealing with semantic gap in microblogging content is more relevant for the quality of recommending like-minded users.