微博舆情事件中用户关系分析技术的研究与实现

发布时间：2019-03-06 18:37

【摘要】：近年来,随着移动互联网的兴起,微博类在线社交应用得到了快速发展,微博所具有的开放性高、互动性强和信息传播迅速等特点使得它迅速演变成了互联网舆情的主要策源地。Twitter和新浪微博作为目前全球使用人数最多,传播面最广的两大微博系统,在舆情分析中占据着重要的地位。本文面向微博舆情分析的需要,从微博用户的自身属性、联系的拓扑结构以及博文的内容分析三个方面入手,总结了显、隐、强、弱四个维度对用户关系的影响程度,通过运用关联规则发现、图挖掘和文本分析等技术手段,来揭示微博用户在事件传播中的社会特性和隐含关系。本文从分析Twitter的认证模式出发,研究了如何突破Twitter限制从而实时高效获取后台数据,并设计了一个Twitter话题圈子分析模型,基于该模型可以对用户圈子以不同属性指标进行划分,并能发现话题圈子中的关键节点和挖掘在话题发展中起主要作用的频繁用户关联模式,最后通过实验证明了该模型在Twitter平台舆情分析中可行且有效。本文使用新浪微博开发平台抓取舆情事件的微博数据,根据博文和话题所表达主题相关性的差异,提出了基于主题相关性分类的微博话题立场研判方法,可以对用户进行立场划分并判断话题的传播立场。此外也研究了话题主题词集和立场研判词库的自动构建方法,在此基础上本文设计了一个基于微博话题立场研判的用户划分模型,可以用于政府有关部门监测互联网舆情以及企业评估产品市场等方面,具有一定的实用价值。本文在最后介绍了面向舆情分析的协同搜索系统整体架构,并对其中人物关联分析模块的功能做了详细介绍。人物关联分析模块主要解决协同搜索任务中的人物推荐排序、话题圈子分析以及话题立场研判等功能,之后对基于二部图的人物和事件关联模型做了简单介绍,随着该模型经过数据积累和功能扩充后,可以不断完善面向互联网舆情分析的实体关联知识图谱,为以后超大规模数据下的舆情分析打下坚实的基础。
[Abstract]:In recent years, with the rise of the mobile Internet, Weibo's online social applications have been rapidly developed, Weibo has a high degree of openness, The characteristics of strong interaction and rapid information dissemination make it rapidly evolve into the main source of Internet public opinion. Twitter and Sina Weibo, as the two largest Weibo systems with the largest number of users and the widest spread in the world, are currently the two largest users in the world. It occupies an important position in the analysis of public opinion. Facing the needs of Weibo's public opinion analysis, this paper summarizes the influence degree of the four dimensions on the user relationship from three aspects: Weibo user's own attributes, the topological structure of the connection and the content analysis of the blog post, and summarizes the influence degree of the four dimensions to the user relationship, namely, explicit, implicit, strong and weak. By means of association rule discovery, graph mining and text analysis, this paper reveals the social characteristics and implicit relationship of Weibo users in the event propagation. Based on the analysis of the authentication mode of Twitter, this paper studies how to break through the limitation of Twitter to obtain the background data efficiently in real time, and designs a Twitter topic circle analysis model. Based on this model, the user circle can be divided into different attribute indexes. The key nodes in the topic circle and the frequent user association pattern which play a main role in the topic development can be found. Finally, the experiment proves that the model is feasible and effective in the public opinion analysis of Twitter platform. This article uses Sina Weibo development platform to grab Weibo data of public opinion events, according to the difference of topic relevance expressed in blog post and topic, puts forward Weibo topic position research method based on theme correlation classification. The user can divide the position and judge the communication position of the topic. In addition, this paper also studies the automatic construction method of topic thesaurus and position judgment thesaurus. On this basis, this paper designs a user partition model based on Weibo's topic position judgment. It can be used to monitor the public opinion on the Internet and evaluate the market of the products by the relevant government departments, which has certain practical value. At the end of this paper, the architecture of collaborative search system for public opinion analysis is introduced, and the function of character association analysis module is introduced in detail. Personage correlation analysis module mainly deals with the functions of character recommendation sorting, topic circle analysis and topic position analysis in collaborative search tasks, and then gives a brief introduction to the character and event association model based on bipartite graph. After data accumulation and function expansion, the model can continuously improve the entity association knowledge graph for Internet public opinion analysis, and lay a solid foundation for future public opinion analysis under super-large-scale data.
【学位授予单位】：国防科学技术大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.092;TP391.3

【参考文献】