社交媒体事件检测研究综述  

Survey on Event Detection Research in Social Media

作　　者：王冰玉 吴振宇[1] 沈苏彬[2] 陈佳颖 WANG Bing-yu;WU Zhen-yu;SHEN Su-bin;CHEN Jia-ying(School of Internet of Things,Nanjing University of Posts and Telecommunications,Nanjing 210000,China;School of Computer Science ＆Technology,Nanjing University of Posts and Telecommunications,Nanjing 210000,China)

机构地区：[1]南京邮电大学物联网学院,江苏南京210000 [2]南京邮电大学计算机学院,江苏南京210000

出　　处：《计算机技术与发展》2018年第9期105-111,共7页Computer Technology and Development

基　　金：国家自然科学基金青年项目(61502246);南京邮电大学科研启动基金项目(NY215019)

摘　　要：事件检测是社交媒体挖掘的重要内容之一。目前,已经提出多种针对社交媒体数据的事件检测方法。然而,对事件的定义以及检测方法的优缺点尚未明确说明。因此,首先对事件的定义进行了说明,分析了事件与话题等易混淆概念之间的区别与联系,事件相对于话题更具局限性,而同一话题下可能涵盖多个相似或相关事件。其次,从社交媒体数据类型的角度出发,分析和总结了社交媒体事件检测方法的优缺点以及适用场景,传统媒体中常使用的基于Single-Pass、基于突发项等原理及实现简单,但是适用场景具有局限性,基于聚类的方法可实现无监督的事件自动检测,但其大部分实现都相对复杂。基于社交数据的方式则可以利用用户行为信息更及时地发现热点事件。最后,对事件检测的未来发展方向进行了展望。Event detection is an important part in social media data mining field. At present,a variety of methods of event detection to so-cial media data are put forward. However,the definition of event and the advantages and disadvantages of exist methods have not beenclearly stated. Therefore,we firstly describe the definition of event and analyze the differences and connections between confusing con-cepts such as events and topics. In general,the concept of event is narrow than topic as the same topic may cover multiple similar or relat-ed events. Secondly,from the point of data type,the merits and demerits of event detection methods and the application scenarios are dis-cussed and summarized. Methods based on Single-Pass and burst items are simple and easily implemented while having the limitations ofapplication scenario. Clustering-based methods can detect events without supervision but are relatively complex and time-costing. Meth-ods aiming at social user data is a new direction deserve to deep research as they utilize user behavior information in time which can detectevents more timely. Finally,the future direction of event detection is prospected.

关 键 词：事件检测 事件 话题 社交媒体 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]