基于网络舆情数据的用户属性及事件情感分析

发布时间：2019-01-17 07:15

【摘要】：随着大众媒体的发展与普及,互联网已经成为网络用户获取信息、表达意见、交流看法的重要平台。官方新闻网站发布的信息经过民间社交网络的讨论形成舆论流,影响事件的传播与演化过程。然而信息在以权威媒体为代表的官方舆论场与以个体网络用户为代表的民间舆论场的交互机制尚不明确,进而当突发事件爆发时难以控制不良情感的发生、传播与演化。因此,事件的网络舆情发展肌理已经引起了大量研究者的兴趣。用户在网络中的行为是其在现实世界中行为的映射,用户的内在特征影响事件的信息的传播与演化过程,反应在网络中即为发表言论及其所表达心情的差异。因此,通过挖掘用户在网络中的言论行为以及用户的情感态度能够发现用户的个性化属性,进而能够分析其在事件传播与演化中的角色与作用。在社交网络中,信息经过转发、评论、点赞等方式进行传播,同时若用户在转发与评论信息时添加个人看法,那么便促进了事件的演化。鉴于此,本文结合认知科学与媒体传播学科的研究方法和手段,对网络中用户的基本属性、性格属性,以及信息传播机制、事件情感走势等问题进行了研究。重点讨论了影响网络舆情传播与演化过程的用户模型与事件传播演化模型,并对用户的背景情感知识进行计算表达,从微观与宏观两个方面对用户的信息接受与传播机制进行研究。论文的主要工作和创新点如下:1.基于社会心理学对情感分析的结论,提出基于模糊逻辑的多维多粒度情感计算方法。该方法依赖具体语境,计算词汇在语境中的情感,同时自适应加入网络流行用语。此外,基于模糊逻辑的方法使得词汇或者微博在各个维的隶属度均为[0,1],打破其加和为1的限制,这一点使得情感的模糊计算更加符合人类的认知过程。2.基于社交网络相关数据,建立用户模型。该用户模型考虑用户的基本属性,包括用户的性别、年龄、教育水平,并对缺失基本属性的用户信息进行补全。此外该用户模型包括用户内容背景知识、用户兴趣点、情感背景知识、个性化性格特征等,挖掘用户在接收信息与传播信息的个性化特征。3.提出基于官方舆论场与民间舆论场的事件情感交互分析方法。该方法不仅对事件的内容分析,而且对事件的情感进行度量,并进一步提取官方舆论场与民间舆论场的诱导关系。该方法能够辅助预测事件发展趋势,进而可以为调控事件提出决策建议。4.基于网络动力学,研究用户在信息传播时的级联现象。在用户联系紧密的社区内部,容易形成信息传播的级联现象;在社区之间,用户之间联系疏松时,容易形成信息演化现象。一旦形成级联,事件的原始信息便发生丢失,出现“随大流”现象。
[Abstract]:With the development and popularization of mass media, the Internet has become an important platform for network users to obtain information, express opinions and exchange views. The information released by the official news website is discussed by the folk social network to form the public opinion flow, which affects the process of the event's propagation and evolution. However, the interaction mechanism of information between the official public opinion field represented by the authoritative media and the folk public opinion field represented by individual network users is not clear, and it is difficult to control the occurrence, dissemination and evolution of bad emotions when emergencies break out. Therefore, the development of network public opinion has attracted the interest of a large number of researchers. The behavior of users in the network is the mapping of their behavior in the real world. The inherent characteristics of users affect the process of information dissemination and evolution of events. Therefore, by mining the user's speech behavior in the network and the user's emotional attitude, we can discover the personalized attributes of the user, and then analyze its role and role in the event propagation and evolution. In social networks, the information is transmitted by forwarding, commenting and liking, and if users add their own opinions when forwarding and commenting on the information, it will promote the evolution of events. In view of this, combined with the research methods and means of cognitive science and media communication, this paper studies the basic attributes, personality attributes, information communication mechanism and event emotion trend of users in the network. The user model and event propagation evolution model, which affect the process of network public opinion propagation and evolution, are discussed, and the background emotional knowledge of users is calculated and expressed. This paper studies the mechanism of information acceptance and dissemination from micro and macro aspects. The main work and innovation are as follows: 1. Based on the conclusion of affective analysis in social psychology, a multi-dimensional and multi-granularity emotional calculation method based on fuzzy logic is proposed. The method relies on specific context, calculates the emotion of vocabulary in context, and adaptively adds popular online terms. In addition, the fuzzy logic based approach makes the membership of vocabulary or Weibo in each dimension [0 / 1], breaking the limit of the sum of 1, which makes the fuzzy calculation of emotion more in line with the cognitive process of human beings. 2. Based on the relevant data of social network, the user model is established. The user model considers the basic attributes of the user, including the user's gender, age and education level, and complements the user information that lacks the basic attribute. In addition, the user model includes user content background knowledge, user interest point, emotional background knowledge, personalized personality features, etc. A method of event emotion interaction analysis based on official public opinion field and folk public opinion field is proposed. This method not only analyzes the content of the event, but also measures the emotion of the event, and further extracts the inductive relationship between the official public opinion field and the public opinion field. This method can help to predict the trend of event development, and then can make decision suggestions for the control of events. 4. 4. Based on network dynamics, the concatenation phenomenon of users in the process of information transmission is studied. It is easy to form cascading phenomenon of information dissemination in the community where users are closely connected, and the phenomenon of information evolution is easy to form when the connection between communities is loose. Once the cascade is formed, the original information of the event is lost and the phenomenon of "following the stream" occurs.
【学位授予单位】：上海大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP393.09

【共引文献】