Discovering Twitter Users' Off-line Community

Twitter is a fast-growing online social network service (SNS) where users can "follow" any other user to receive his or her mini-blogs which are called "tweets" Two unique features distinguish Twitter from other SNS:(Ⅰ)The absence of mutual consent in establishing follow links; and (Ⅱ) Being a mixture of news media and social network. To better understand Twitter user behavior, an important question is how much of this Twitter network reflects one's real-life social network. In this paper, we study the problem of identifying a user's off-line real-life social community, which we call the user's Twitter off-line community, purely from examining Twitter network structure. Based on observations from our user-verified Twitter data and results from previous works, we propose three principles about Twitter off-line communities. Incorporating these principles, we develop a novel algorithm to iteratively discover the Twitter off-line community based on a new way of measuring user closeness. According to ground truth provided by real Twitter users, our results demonstrate the effectiveness of our approach with both high precision and recall in most cases.