基于朴素贝叶斯的网络舆情话语生态分析

Ecological Analysis of Network Public Opinion Discourse Based on Naive Bayes

进入互联网时代,各网络平台逐渐成为人们述说己见(网评)的重要阵地,而这些网评时时夹杂着各种情绪偏激、内容低俗等语言信息,从语言生态学角度来分析,这是典型的话语生态失衡的表现。对此相关网络监管部门有必要实时监督,及时处理,进行反馈并加以引导。文章以弹幕评论这一新兴的短小语言信息为研究对象,利用网络爬虫数据挖掘技术采集B站弹幕评论数据,用Pandas库和正则表达式等技术完成数据清洗,在朴素贝叶斯分类基础上,利用Jieba库实现分词,运用情感分析进行研判,得到舆情话语生态分析的主要语料,在此基础上筛选出负面情绪的弹幕语料进行生态话语分析。实验基于朴素贝叶斯对案例内容分类并加以训练,可得到评价分类与预测结果,实验表明,该系统稳定可靠,具有良好的效果。

In the Internet Age,various online platforms have gradually become an important position for people to express their opinions(online reviews).These online reviews are always mixed with various language information such as extreme emotions and vulgar content.From the perspective of language ecology,this is a typical expression of discourse ecological imbalance.The relevant network regulatory authorities should supervise and handle this in a timely manner,and then provide feedback and guidance.This paper takes the bullet screen comment,a new short language information,as the research object,and uses web crawler data mining technology to collect the bullet screen comment data of bilibili,and then uses Pandas database and regular expression technology to complete data cleansing.On the basis of naive Bayesian classification,Jieba database is used to implement word segmentation,and emotional analysis is used to study and judge,and the main corpus of ecological analysis of public opinion discourse is obtained,Based on this,select bullet screen corpus of negative emotions for ecological discourse analysis.This experiment is based on naive Bayesian classification of case content and training,which can obtain evaluation classification and prediction results.The experiment shows that the system is stable,reliable,and has good results.