A Case Study of Mining and Correlation Analysis of Public Security Events in Heterogeneous and Unstructured Web Messages
Incidents of public security have an ascendant trend in recent years all over the world, and it is more important to understand the correlation of different kinds of public security incidents.With the popularization of the Internet, numerous web messages can provide resources to do that.However, an important challenge is that the web messages are often heterogeneous and unstructured.In this paper, we propose a methodology to extract events from noisy web messages automatically.Then, based on the essence and the propagation features of event ontologies, we analyze the correlation between three typical kinds of public security incidents in China.With a heterogeneous dataset containing microblog messages and news reports during 2011-2014, the case study shows that our methodology can locate the security event on timeline with an error less than 1 day.Furthermore, the analysis from multiple views not only reveals the correlation between different kinds of public security event, but also reveals the difference between the heterogeneous resources of news report and microblog.
Event extraction Public security Heterogeneous and unstructured Social media
Yinghui Wang Bo Wang Xuelun Li
School of Computer Science and Technology Tianjin University Tianjin, China Qiushi Honors College Tianjin University Tianjin, China
国际会议
武汉
英文
115-120
2016-09-23（万方平台首次上网日期，不代表论文的发表时间）