Abstract—Microblog has witnessed an explosive development in the past few years. The research on how to detect bursty events timely and efficiently from large amounts of microblog posts has attracted much attention. The key to bursty events detection is the extraction of bursty features. We reconsider the problem from the standpoint of Kinematics theory and look on bursty events detection as a Kinematics phenomenon instead of considering bursty features separately. A novel model named PHCM (Particle Horizontal Cast Movement) is proposed in this paper. Bursty words describing the particular event tend to appear simultaneously in microblog streams. So we shuffle the bursty words appearance sequence in time scale and sort them in ascending order by frequency. Each bursty word is regarded as a moving particle over the time dimension and the movement can be decomposed into a series of consecutive flat parabolic motions with specific time interval. Then the particle’s vertical velocity and horizontal velocity in each interval are taken to construct the feature vector of its trajectory. Finally, through computing similarity between different trajectory vectors, we can get bursty words clusters with similar trajectories. By retrieving microblog posts containing bursty words in the same cluster, bursty events can be obtained. Experiments show that our model can detect bursty events accurately as well as efficiently.

Index Terms—Microblog, bursty events, Kinematics theory, moving particle, trajectories.

Le Zhang and Xueqiang Lv are with Beijing Information Science and Technology University, China.
Leihan Zhang is with Beihang University, China.

Cite:Le Zhang, Xueqiang Lv, Leihan Zhang, "PHCM: A Particle Horizontal Cast Movement Based Model for Bursty Events Detection of Chinese Microblog," International Journal of Computer and Communication Engineering vol. 4, no. 6, pp. 426-438, 2015.