网络舆情语义识别的技术分析及识别流程构建

发布时间：2018-01-07 00:38

本文关键词：网络舆情语义识别的技术分析及识别流程构建　出处：《吉林大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着现代社会网络化进程的加速,网络已经成为人们表达个人意见和观点的重要渠道。网络舆情信息成为舆情信息的一个关键部分。网络舆情和传统舆情相比,具有数据量大、突发性强、影响范围广等特点。在大数据环境背景下,网络舆情信息的挖掘较以往的传统媒体更加困难,并且更加难以规范。因此,如何利用语义识别技术,有效地从海量的舆情信息中挖掘出其中的关键因素并指导决策以及突发事件的处理就成为舆情研究的重要方向。本文通过分析研究网络舆情语义识别的各类技术,重新建立一套网络舆情语义识别的流程。该流程不仅包含网络舆情语义识别的技术细节,同时也建立起一套网络舆情突发事件的评判标准提供参考。本文所构建的网络舆情语义识别的流程包含网络舆情信息采集模块、网络舆情预处理模块、网络舆情话题识别模块以及网络舆情反馈模块等四组子模块,前三个模块的主要工作是对网络舆情进行识别分析,从浩如烟海的网络信息中将网络舆情信息加以提取,所以这三个子模块是舆情预警与疏导的基础;而网络舆情反馈模块则是基于上述三个子模块的识别结果对网络舆情态势进行系统的分析,并针对舆情突发事件的不同爆发状态提出相应的疏导策略。本文针对当前网络舆情爆发的主要阵地微博平台提出了实证研究,选取了当前的舆论热点话题进行语义识别分析,并预警突发事件,提出疏导建议。本文的创新点主要有:将网络舆情信息语义识别技术进行归类,比较分析各种语义识别技术找到其各方面的特征,比较分析的方面主要有:信息处理的精度比较、人工参与程度比较、特征库比较、多媒体信息的处理能力比较、深层次语义信息的挖掘比较、技术复杂程度比较以及通用性、适应性的比较。对网络舆情信息的语义识别技术进行研究,并根据网络舆情的特点,提出一个网络舆情语义识别的技术流程方案。整个流程分为信息采集、预处理、话题分析、舆情反馈等几个步骤。在网络舆情信息的采集过程中,本文采用通用网络爬虫技术对网络信息进行爬取;在获取了基础的数据集合之后,需要对信息集合进行预处理,将文本进行分析处理并去除文本中的停用词,对文本信息的特征进行抽取;随后,对舆情信息的文本集合进行聚类分析,使用文本向量模型对文本信息金星表示,采用K-means聚类算法对文本信息进行聚类,挖掘信息话题;对信息的情感倾向性进行分提取,对其情感倾向的强弱进行排序,得出舆情语义识别的结果。
[Abstract]:With the acceleration of modern social network, the network has become an important channel for people to express personal opinions and views. The network public opinion information has become a key part of public opinion information. Compared with traditional network public opinion and public opinion, with a large amount of data, strong burst characteristics, affecting a wide range. In the big data environment, mining network public opinion information than the previous traditional media is more and more difficult, and more difficult to regulate. Therefore, how to use semantic recognition technology, effectively from massive public opinion information to dig out the key factors and guide the decision-making and handling of emergencies has become an important direction of public opinion research. Through the analysis of network public opinion research in semantic recognition all kinds of technology, to establish a set of network public opinion semantic recognition process. The process includes not only the technical details of the semantic recognition of network public opinion, also built Set up a set of reference network publicsentiment emergency evaluation standard. The network public opinion semantic recognition process includes information collection module, preprocessing module of network public opinion, the network public opinion, the network public opinion topic identification module and network public opinion feedback module of four sub modules, the main work of the first three modules is analyze public opinion on the network, to extract information from the network, the network public opinion information multitude, so the three sub module is the basis of public opinion warning and persuasion; and the network public opinion feedback module is the identification of the three sub modules based on the results of the online public opinion situation are systematically analyzed, and according to the different publicsentiment emergency outbreak status of leading the corresponding strategies. Aiming at the main position of the current network outbreak of public opinion presents an empirical study on the micro-blog platform, select the Public opinion topic semantic recognition analysis, early warning and emergency, put forward guidance suggestions. The main innovations of this paper are: the network public opinion information semantic recognition technology classification, comparative analysis of various semantic recognition technology to find the characteristics of the various aspects of the comparative analysis are: comparison of the accuracy of information processing, artificial participation degree feature library comparison, comparison, processing of multimedia information, mining deep semantic information, technical complexity and versatility, adaptability. Research on semantic recognition technology of network public opinion information, and according to the characteristics of network public opinion, put forward a network public opinion semantic recognition technology. The whole process flow scheme is divided into information collection, preprocessing, topic analysis, public opinion feedback and other steps. In the process of collecting information in network public opinion, the general network Network crawler technology for network information crawling; after obtaining the basic data set, the need for information collection pretreatment, text analysis and text processing to remove the stop words, to extract features of text information; then, the collection of information for cluster analysis, said the text use the text information of Venus vector model, K-means algorithm is used to cluster the text information mining, information extraction of topic; emotional tendency of information, on the strength of the emotional inclinations that sort of public opinion semantic recognition results.

【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：G254

【参考文献】