【作者】 董坚峰；

【导师】 张玉峰；

【作者基本信息】 武汉大学 ， 情报学， 2013， 博士

【摘要】 20世纪末的信息革命推动了现代网络技术的广泛应用和普及,网络媒体作为“第四类舆情传播媒体”逐步受到关注,并成为舆情信息汇集的新通道和“观点的集散地”进入主流社会视野。尤其是近年来,中国网民数量逐年递增,网民借助互联网平台进行利益诉求表达、情绪宣泄、公共事务参与的积极性和频率愈来愈高,国家在社会转型期间凸现的各类社会矛盾也往往在第一时间聚焦于网络平台,形成强大的舆论场。与此同时,互联网强大的开放、快速传播和互动特性使得一些原本在传统传播渠道和言论环境下简单可控的舆情事件转化为群体性突发事件,引发社会公共危机,并给政府等公共管理部门带来了极大的压力和挑战。作为政府公共管理的一项重要职能,网络舆情危机预警近年来受到广泛关注,与此相关的机制建设和技术系统研制纷纷展开,网络舆情监控和预警受到了包括国家层面在内的各级政府重视。然而,作为网络舆情危机预警流程中最关键的一环,网络舆情的分析和研判相关的研究和实践远远滞后于当前Web2.0环境下的公共危机预警现实需求,无论在舆情分析方法还是在技术支撑方面均无法提供对当前动态、交互的网络舆情进行自动化、智能化和实时化的危机预警支持。在这种背景下,利用现有技术手段和学科方法对网络舆情分析方法和技术进行创新进而为危机预警提供技术保障,对维护社会稳定、推进社会转型具有重要的现实意义。本文正是从这一实际出发,以网络舆情这一特殊的信息客体为研究对象,在梳理前人研究成果的基础之上,将网络舆情分析与公共危机预警目标相结合,综合采用信息分析、自然语言处理、人工智能、数据挖掘、危机管理、新闻传播等多学科理论与方法,从多个角度和层面对网络舆情分析的理论、方法、技术和应用进行研究。本论文共包括七个部分,主要内容如下(1)引言本部分介绍了论文的选题背景和选题目的,利用文献调研等方法对与本论文相关的国内外研究现状进行了梳理,为后续研究奠定了理论积累。(2)公共危机中的网络舆情研究本部分首先从分析危机和公共危机的内涵出发,探讨了当前公共危机的特征及发展趋势；同时,深入剖析了网络舆情的概念内涵、构成要素、生命周期。在此基础上,对公共危机中的网络舆情的素材来源、特征表现、内容表征和传播演变规律进行了探讨,并以邓玉娇事件为例分析了网络舆情传播演变规律。(3)面向公共危机预警的网络舆情分析框架构建本部分从剖析当前网络舆情公共危机预警严峻的现状入手,分析了网络舆情分析的必要性和现实需求；同时,对公共危机预警的流程和功能进行了分析,结合危机预警需求构建了面向公共危机预警的网络舆情分析框架,该框架主要由舆情信息源、技术支撑、分析方法、分析模式、分析与应用五层构成,基本涵盖了网络舆情分析涉及的从技术到方法的各层面内容。在舆情分析流程方面,借鉴传统的信息分析流程提出了“舆情主题规划——舆情信息收集——舆情信息预处理——舆情信息分析——舆情危机预警处理”流程,并阐述了分析要求和原则。本部分最后系统阐述了网络舆情分析中涉及的自然语言处理、舆情信息采集与特征提取、语义挖掘和文本倾向性分析等关键技术,为分析方法的实施和技术系统的构建提供了技术支撑。(4)面向公共危机预警的网络舆情分析方法研究本部分主要探讨了网络舆情分析的方法论。首先简要介绍了网络舆情分析中用到的各种方法,并构建了网络舆情分析方法体系。随后对网络舆情分析涉及的内容分析法、网络计量法和智能分析方法等重要方法进行了详细阐释。内容分析法和网络计量法是社会科学研究中普遍使用的科学方法,在对网络舆情这一新型信息客体进行分析时仍然适用。在智能分析方法方面,本文提出了建立在不同技术基础之上的两类不同的智能分析方法,其中基于Web挖掘的方法是数据挖掘在网络信息分析中的新应用,目前已经成为主流的舆情分析方法,该方法主要是从Web文本内容、Web链接结构、Web使用记录等方面发现舆情规律,包括Web文本挖掘法、主题聚类法、链接挖掘法和社会网络分析法等方法；基于语义的方法则主要通过分析Web文本中潜在的语义结构或借助本体、语义词典等外部语义知识从舆情文本语义层面发现舆情规律,主要包括潜在语义分析法和基于外部语义知识的语义分析方法等。(5)面向公共危机预警的网络舆情分析模式研究本部分以在公共危机预警中起关键作用的四个网络舆情要素作为分析对象,分别从话题发现、事件识别、网民个体分析、群体行为分析四个方面探讨网络舆情分析的主要模式。在进行话题发现模式分析时,从话题发现的流程入手,探讨了话题建模和话题趋势分析的方法,然后对热点话题和敏感话题这两个引发舆情危机的关键点的舆情发现及影响评估的方法进行了重点探讨。在进行事件识别模式分析时,探讨了公共危机事件要素提取的方法,在进行事件建模的基础上分析了突发事件的识别过程,最后对舆情事件危机程度及其影响分类进行了研究。在网民个体分析模式分析方面,从分析公共危机舆情中的网民构成及特征入手,探讨了如何借助文本倾向分析、链接分析等方法实现对网民个体的情绪分析和个体影响力分析,最后以网络意见领袖这一特殊舆情主体为对象探讨了舆情意见领袖识别方法。在进行群体行为特征及行为模式分析时,以网络社群这一主要网络舆情群体为对象,分析了他们在网络舆情公共危机中的群体行为演化规律及呈现的特征,并利用社会网络分析法、趋势分析法等方法对网络舆情公共危机中的群体行为识别和发展趋势预测等进行了分析。(6)面向公共危机预警的网络舆情分析应用研究本部分主要将理论研究应用于实践,介绍了网络舆情分析的应用示例。应用平台采用了国内拓尔思公司的TRS网络舆情监测系统,实现网络舆情分析和公共危机预警。文中详细讨论和演示了该系统从舆情信息采集、预处理、舆情智能挖掘和分析到预警服务提供的全过程。(7)总结与展望本部分首先从论文研究意义和完成的工作等方面对全文进行总结。然后,反思了论文存在的不足,并展望了后续研究工作的重点。论文最后指出,随着国内外学术界对网络舆情分析关注的不断深温,在计算机、信息管理、公共管理和新闻传播等多领域专家学者和从业人员的推动与努力下,网络舆情分析研究将会迎来新一轮的发展高潮。本文系国家自然科学基金项目“基于数据挖掘的企业竞争情报智能采集机制研究”(编号：70573082)和“企业竞争情报智能分析模型与方法研究”(编号：71073121)、教育部人文社会科学重点研究基地重大项目“基于智能信息处理的知识挖掘技术及应用研究”(编号：08JJD870225)和湖南省教育厅科学研究项目“基于数据挖掘的网络舆情预警机制研究(编号：11C1051)的研究成果。更多还原

【Abstract】 With the propelling of information revolution in the end of the20th century, modern network technique has been widely applied and popularized, and online public media, as the fourth kind of communication media for public opinion after newspaper, broadcast and television, has gradually received attention and entered into the mainstream society’s viewpoint for the role of new channel for the collection of public information and distributing centre of public opinion. Especially in the recent years, with the progressive increase of the quantity of Chinese netizens year by year, more and more netizens use Internet to express their interest appeal, abreact their emotion, participate public affairs, and all kinds of social contradictions producing in the social transformation period are always focusing on the online public platform to form powerful public opinion field at the first time. Meanwhile, the strong characteristics of openness, rapid spread and interaction for Internet make some simple and controllable public opinion incidents in the traditional communication channels and opinion environment turn to unexpected group events and trigger social public crisis, which bring huge pressure and challenge to public administration.As an important function for public administration, online public opinion crisis pre-warning has been paid extensive attention by the public, and the relevant mechanism construction and technical system development has proceeded one after another, governments at all levels pay more attention to the monitoring and pre-warning of public opinion. However, the analysis and judging of online public opinion, as the key link of crisis pre-warning processes, whose relevant research and practice falls behind the practical demand for public crisis pre-warning under web2.0environment, and can’t provide automatic, intelligent and real-time crisis pre-warning support to current dynamic and interactive online public opinion from the aspects of analysis methods and technical support. In this context, using current technical means and disciplinary methods to innovate methods and technology of online public opinion analysis to provide technical guarantee for crisis pre-warning has important practical significance for maintaining social stability and propelling social transformation.This paper starts from the above reality, taking online public opinion as research object, teasing the former scholars’research achievement as the research basis, combining online public opinion analysis and the target of public crisis pre-warning, and using the theory and methods of information analysis, natural language processing, artificial intelligence, data mining, crisis management, journalism and communication, etc to study the theory, methodology, technology and application of online public opinion analysis from various aspects and levels. The paper includes seven parts, and the main idea of every part lists as the following:(1) IntroductionIn this part, the author introduces the background and research objective of this paper, and teases the research status relevant to this paper to establish the theoretical basis for the follow-up study.(2) Study on the online public opinion in public crisisIn this part, the paper starts from analyzing the connotation of crisis and public crisis, discussing the characteristics and development tendency of current public crisis. Meanwhile, the paper dissects the material sources, concept connotation, characteristics representation, content expression, spread and evolution rule of online public opinion, and takes Deng yujiao event as an example to analyze the spread and evolvement rule of online public opinion.(3) The construction of analysis framework of online public opinion oriented public crisis pre-warningIn this part, the paper starts from analyzing the rigorous status of public crisis pre-warning for online public opinion, demonstrating the necessity and realistic demand. Meanwhile, it analyzes the processes and functions of public crisis pre-warning, constructing the analysis framework of online public opinion oriented public crisis pre-warning, which constituting information sources layer, technical support layer, analysis method layer, analysis pattern layer, analysis and application layer, covering all the aspects of online public opinion analysis from technical aspect to methodological aspect. In the respective of online public opinion analysis processes, the paper learns from the traditional information analysis processes and puts forward the analysis processes including theme planning, information collecting, information preprocessing, information analysis and crisis per-warning and processing, and expounds the analysis demands and principles. In the end, the paper demonstrates the relevant technologies of online public opinion analysis, taking natural language processing, information collection and feature extraction, semantic mining, text orientation analysis, etc as key technologies to provide technical support for the implement of analytical method and establishment of technology system.(4) Study on the analysis methods of online public opinion oriented public crisis pre-warningIn this part, the paper mainly discusses the methodology of online public opinion analysis. First, the paper briefly introduces various important methods using in online public opinion analysis and constructs the method system of online public opinion analysis. Then the paper demonstrates content analysis method, webometrics method, and intelligent analysis method relevant to online public opinion analysis in detail. Content analysis and webometrics are the general scientific methods using in social sciences research, and still suitable for analyzing online public opinion. As to intelligent analysis method, the paper puts forward two different kinds of intelligent analysis methods based on different technologies, one is based on web mining, and another is based on semantic analysis. The analysis method based on web mining is the new application of data mining in online public information analysis, focusing on discovering online public opinion law from the content of web texts, the structure of web links and the usage records of website, including such methods as web text mining, topic clustering, link mining and social online public analysis, and it has become the mainstream method for opinion analysis. The analysis method based on semantic focuses on discovering opinion law from the semantic layer assisted by analyzing the latent semantic structure in web text or ontology and semantic dictionary, including latent semantic analysis and semantic analysis based on external semantic knowledge, etc.(5) Study on the analysis patterns of online public opinion oriented public crisis pre-warningIn this part, the paper takes four key opinion elements acting in public crisis pre-warning as analysis object, discussing the main analysis patterns from the aspects of topic discovering, event recognizing, individual netizen analysis and group behaviors analysis. As to the analysis of topic discovering pattern, the paper starts from the processes of topic discovering, discussing the method of topic modeling and developing trend. Then the paper emphasizes on hot topic and sensitive topic, the two key factors of triggering public crisis, and discusses the method of opinion discovering and impact assessing. As to the analysis of event recognizing pattern, the paper discusses the method of extracting the elements of public crisis event. Based on event modeling, the paper analyzes the recognizing processes of emergent event and the assessment of its influence degree and the classification method. As to the analysis pattern of individual netizen, the paper starts from analyzing the constituent and characteristics of netizen in online public opinion crisis, probing how to use text tendency analysis method, link analysis method to analyze the emotion and influence for individual netizen, then discusses the recognizing method for the opinion leader, a special opinion subject in online public opinion crisis. As to the analysis of the behavior characteristics and behavior pattern for Internet population, the paper takes online public community as the main opinion object and analyzes the evolution law and characteristics of the online public community. Meanwhile, the paper discusses the method for recognizing group behavior and predicting their development trend by social online public analysis and trend analysis.(6) Study on application of the online public opinion analysis oriented public crisis pre-warningIn this part, the paper applies the theory of opinion analysis to the practice and introduces an analysis example based on TRS opinion management system. This system is developed by TRS Company, and can finish the work of online public opinion analysis and public crisis pre-warning. Then the paper demonstrates the whole processes for TRS OM to carry out information collection, information pre-processing, intelligent mining and analyzing, pre-warning service providing in detail.(7) Conclusion and further studyIn this part, the author firstly concludes the whole paper from research significance and the finished work. Then, the author introspect the weakness existing in the paper and outlooks the key points for the further research. In the end, the paper states that with more and more domestic and overseas scholars emphasizing the study of online public opinion analysis, the research of online public opinion analysis will welcome a new round of developing climax by the promotion and efforts of the experts and professionals from the fields of computer science, information management, public administration and journalism&communication.This paper has been funded by the National Natural Science Foundation of China under grant70573082and71073121, the MOE Project of Key Research Institute of Humanities and Social Sciences in Universities under grant08JJD870225and Scientific Research Fund of Hunan Provincial Education Department under Grant11C1051.更多还原