Abstract: Nowadays it can be observed that many companies getting involved in public events on the Internet generally employ public relations company to flood the Internet with purposeful articles or comments. Often hired by public relationship company, the online paid posters earn money by posting comments and articles in forum or social media. Their attempts can effectively impact on public opinions toward the involved companies. In China, these online paid posters are called water army. This paper defines the concepts of emotional social network and factions, and proposes a framework to discover the overlapping community from the emotional social network and detect factions in communities. This study collects comments in Weibo and constructs emotional social network based on the comments. Detecting communities and factions, this study can discover water army concealing in the social media.

Key words: social media, water army, emotional social network, group behavior

摘要： 社交媒体是当下非常热门的一种互联网应用。在发生公共事件时，许多当事人雇佣公关公司在社交媒体上发布虚假信息来影响舆情。这部分为公关公司所服务的社交媒体账号称为水军。水军在当下的中国社交媒体中有泛滥的趋势。基于水军的群体行为分析社交媒体中的水军集团。定义了情感社会网络和情感社会网络中的团的概念；设计了一个分析框架，包括提出了情感社会网络中发现社区，以及从社区中发现对立的派别（称为团）的算法。从微博的评论数据构建情感社会网络，并发现情感社会网络中的团，可以识别水军，而且可以考察社交媒体中的水军集团的特性。采用人工标注的数据集进行了实验。

关键词: 社交媒体, 水军, 情感社会网络, 群体行为