Identification of collective viewpoints on microblogs

Towards hot events, microblogs usually collect diverse and abundant thoughts, comments and opinions from various viewpoints in a short period. In this paper, we aim to identify collective viewpoints from massive messages. Since individuals may have multiple viewpoints on a given event, and individual viewpoints may also change as time goes by, these present a challenge of extracting collective viewpoints. To address this, we propose a Term–Tweet–User (TWU) graph, which simultaneously incorporates text content, temporal information and community structure, to model postings over time. Based on such model, we propose Time-Sensitive Random Walk (TSRW) to effectively measure the relevance between pairs of terms through considering temporal aspects, and then group terms into collective viewpoints. Additionally, we propose Incremental RandomWalk method to recompute relevance between nodes incrementally and efficiently. Finally, we evaluate our approaches on a real dataset collected from Sina microblog, which is the biggest microblog in China. Extensive experiments show the effectiveness and efficiency of our algorithms.