基于BERT模型的网络舆情情感分析——以上海疫情为例

Sentiment Analysis of Network Public Opinion Based on BERT Model—Taking the Shanghai Epidemic as an Example

摘要: 随着互联网时代的不断演进，各种社交软件为网络舆情的传播提供了主要平台。疫情发生后对网络舆情信息的搜集、分析和引导，对于相关部门开展舆情解析和稳定公众情绪具有现实意义。本研究基于新浪微博数据，以此次上海疫情为例，对疫情期间的微博相关数据信息进行提取和挖掘，首先用TF-IDF词频统计处理方法统计出关键节点日期的主题词，然后使用BERT模型对此次疫情的舆情情感进行分析。结果表明：在时间分布上，博文的发布数量整体呈“n”字形分布。在舆情情感上，负面感情前期一直占了较大的比例，中期无明显变化，后期民众的感情虽依旧起伏不定，但消极情绪总体上有了明显的下降趋势。本文研究结果可为日后疫情防控中网络舆情的引导和处置提供借鉴参考。

Abstract: With the continuous evolution of the Internet era, various social softwares provide the main plat-form for the dissemination of network public opinion. The collection, analysis and guidance of online public opinion information after the outbreak of the COVID-19 is of practical significance for rele-vant departments to carry out public opinion analysis and stabilize public sentiment. This study is based on Sina Weibo data, taking the Shanghai epidemic as an example, to extract and mine Weibo-related data information during the epidemic. First, the TF-IDF word frequency statistical processing method is used to count the subject words of the key node dates, then use the BERT model to analyze the public sentiment of the COVID-19. The results show that: in terms of time dis-tribution, the number of blog posts is distributed in the shape of “n” as a whole. In terms of public opinion and emotion, negative emotions have always accounted for a large proportion in the early stage, and there was no significant change in the mid-term. Although the people’s emotions in the later period still fluctuated, negative emotions showed a clear downward trend in general. The re-sults of this paper can provide reference for the guidance and disposal of network public opinion in the future epidemic prevention and control.