网络社会，通过网络表达个人观点与诉求已成为广大民众参与国家事务与政治生活的重要组成部分，网络舆情已成社会热点。作为网络舆情生态环境中最活跃的群体，意见领袖受到了广泛关注。意见领袖识别与网络與情演变模式挖掘因此成为重要研究课题。现有意见领袖识别主要基于外部特征，忽略了帖文本身的内容。本项目将引入情感倾向性分析度量指标，并基于网络文本句法模式分析，提出多维意见领袖识别算法。在网络舆情演化模式分析方面，现有研究绝大部分关注理论与对策以及单个案例的静态分析，缺乏动态观察以及对未来发展趋势的预测。本项目将引入基于Agent的群体仿真理论和方法，关注舆情演变的最活跃主体--意见领袖的行为和认知，形成可预测网络舆情演变路径和态势的模型和规则，为网络舆情的引导和管理对策的制定提供现实依据。

In the network society today, expressing personal views through the Internet has become an important part of life for people to participate in national affairs and political life. Network public opinion has become a hot research topic in China. As the opinion leaders play very important roles in public opinion environment, they have been widely concerned. Thus, identifying the opinion leaders and mining its evolution pattern has become an important issue. . Existing opinion leader identification is based on external characteristics, and has ignored the context of the original posts. In this project, the authors will introduce new context-based metrics of sentiment orientation, analyze the syntax mode of sentences, and identify opinion leaders based on five classification standards. . As to analysis of the evolution model of network public opinion, the existing research focuses on the static specific events instead of necessary dynamic observation and forecast of future trends. In this project, the authors will introduce group simulation theory and methods which focus on opinion leaders' behavior and cognition, form the rules and path of network public opinion's evaluation, and provide a realistic basis for the guidance and management of the network public opinion.