Collective viewpoint identification of low-level participation

Mining microblogs is an important topic which can aid us to gather collective viewpoints on any event. However, user participation is low even for some hot events. Therefore, collective viewpoint discovery of low-level participation is a practical challenge. In this paper, we propose a Term-Retweet-Context (TRC) graph, which simultaneously incorporates text content and retweet context information, to model user retweeting. We first identify representative terms, which constitute collective viewpoints. And then we apply Random Walk on TRC graph to measure the relevance between terms and group them into collective viewpoints. Finally, extensive experiments conducted on real data collected from Sina microblog demonstrated that our proposal outperforms the state-of-the-art approaches.