Abstract:

Bursty events are easy to cause public opinion and are important objects for supervision. Traditional event detection ignores the difference of influence between blog posts. This paper considers the influence of different microblog on event. For the time series Weibo data stream, this paper proposes a bursty events detection framework that combines the influence of microblog and burst words. Based on the comprehensive consideration of the influence of users and blog posts, tap the key microblogs in the window. Calculate burst words based on key microblog and historical data. Then, the latent event dataset with bursty events is constructed by means of burst word retrieval, and the bursty events are detected by clustering algorithm. Compared with two common event detection methods, experiments show that the proposed method has a significant improvement in accuracy and efficiency.

Key words: bursty events, event detection, Weibo, influence of microblog, burst word

摘要：

突发事件容易引起社会舆论，是监管的重要对象。传统事件检测忽略了博文间影响力的差异。考虑到不同微博对事件的影响力不同，针对时序微博数据流，提出一种结合微博影响力与突发词的突发事件检测框架。在综合考虑用户及博文影响力的基础上，挖掘时间窗口内的重点微博，根据重点微博及历史数据计算突发词，再通过突发词检索的方式构建出具有突发性的潜在事件数据集，通过聚类算法检测突发事件。对比两种常见的事件检测方法，实验表明所提方法在准确率与效率上均有明显提升。

关键词: 突发事件, 事件检测, 微博, 博文影响力, 突发词