在线社会网络谣言检测综述  

Review on Rumor Detection of Online Social Networks

作　　者：陈燕方[1] 李志宇[2] 梁循[2] 齐金山 CHEN Yan-Fang;LI Zhi-Yu;LIANG Xun;QI Jin-Shan(School of Inform ation Resource Management,Renmin University of C hina,Beijing 100872;Department of Computer Science,School of Inform ation,Renmin University of China,Beijing 100872)

机构地区：[1]中国人民大学信息资源管理学院,北京100872 [2]中国人民大学信息学院计算机系,北京100872

出　　处：《计算机学报》2018年第7期1648-1677,共30页Chinese Journal of Computers

基　　金：国家自然科学基金(71531012;71271211);北京市自然科学基金(4172032);中国人民大学科学研究基金项目(10XNI029);中国人民大学2017年度拔尖创新人才培育资助计划成果之一资助~~

摘　　要：大数据环境下,在线社会网络与人们的生活、娱乐以及工作逐渐融为一体.然而"信息过载"和"信息污染"已成为在线社会网络诸多应用发展面临的主要瓶颈之一,并同时造成了用户的"信息焦虑"和"信息迷航"等一系列问题,因此在线社会网络谣言检测是改善在线社会网络信息生态环境质量、提升用户体验的有效手段.在线社会网络谣言检测隶属于信息可信度检测研究范畴,但谣言的不确定性、较强的时效性、主观性和关联性等特征又使得其与虚假信息检测有着本质区别.基于以上,该文从在线社会网络谣言的基本概念和特征研究出发,分别基于目标、对象和时间三个属性,分析了在线社会网络谣言检测研究基本问题的形式化定义,并介绍了研究中数据采集和标注的不同方法.然后,分别对不同类别和应用场景的在线社会网络谣言检测方法和谣言源检测方法进行了分析和总结.最后,该文讨论了在线社会网络谣言检测技术未来发展面临的若干挑战以及可能的研究方向.Online Social Networks(OSN)are integrated into people’s life,entertainments,and works with the development of big data.However,the issues of information overload and information pollution have become one of the most serious problems hindering the improvements of many OSN applications,decreasing the cost of misinformation diffusion,and resulting in the widely spreading of rumor information,which may also cause information anxiety and information disorientation among people.Thus,OSN rumor detection is an effective way to improve the quality of OSN information ecology environment and user experience.To some extent,rumor detection belongs to the research area of information credibility evaluation;however,the unique features,including uncertainty,time-effective,subjectivity,and relevance,et al.,of rumor detection make a big difference from misinformation detection.Currently,there are three main problems in rumor detection,namely timeliness,extendibility and scalability,which are great challenges for all the researchers in this field to tackle.In this paper,we firstly summarized some basic concepts(rumor,online rumor and online social network rumor)and the main features of OSN rumor on five perspectives(who,say what,in which channel,to whom,with what effect).And then,we discussed the problem statement of its hot research topic based on different perspectives including detection objective,detection object and detection time to answer the three 3W questions in the process of rumor detection,namely what,who and when.In addition,we showed different ways of rumor data collection(platform,content,quantity,proportion,and method)and data annotation(manual annotation and machine annotation),and some open online rumor datasets.Next,we compared and demonstrated the various performances,strengths,weakness,and applications for both OSN rumor detection based on content and rumor source detection based on network structure techniques.For OSN rumor detection,content classification and content comparison are most common methods curre

关 键 词：在线社会网络 谣言 虚假信息 谣言检测 谣言源检测 网络结构分析 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]