基于情感分析的突发公共卫生事件舆情时空演化差异研究

Spatiotemporal Evolution of Public Opinion in Public Health Emergencies Based on Sentiment Analysis

【目的/意义】分析网民在突发公共卫生事件中的情感演化历程,探究影响网民情感波动的因素及其时空演化的差异。【方法/过程】运用Python爬取微博新冠疫情相关文本数据315 445条,基于SnowNLP情感分析工具对数据文本进行情感分析。使用TF-IDF及LDA主题模型进行建模,对不同阶段及不同群体的舆情时空演化及差异进行内容分析。【结果/结论】网民的情感演化呈现阶段性和群体性差异,尽管整体为积极态势,但疫情上升期为负面情绪集中爆发期;网民群体中受教育程度较低的群体情感波动幅度更大,更容易受到舆论的影响,舆情演化更易极化;中心大城市情感波动相对稳定,而引起其他区域网民消极情绪的往往不是疫情本身,而是由疫情引发的负面舆论;普通网民群体较于高影响力群体在舆情演化阶段的负面情绪更为严重,情感演化在各阶段呈现明显的涟漪效应,需在不同阶段针对不同群体制定有效的舆情引导政策。【创新/局限】本文将整个语料库划分为50多个小语料,个别语料文本数据量较少,具有一定的局限性。

【Purpose/significance】This paper analyzes the emotional evolution process of netizens in public health emergencies,and explores the factors that affect the emotional fluctuation of netizens and the differences in space-time evolution.【Method/process】Python was used to crawl 315445 pieces of COVID-19 related text data from Micro-blog,and sentiment analysis was performed on the data text based on SnowNLP sentiment analysis tool.The TF-IDF and LDA theme models were used for modeling,and the temporal and spatial evolution and differences of public opinion in different stages and groups were analyzed.【Result/conclusion】The emotional evolution of Internet users showed phased and group differences.Although the overall situation was positive,the rising period of the epidemic was the period of concentrated outbreak of negative emotions;Among the netizens,the group with lower education level has greater emotional fluctuation,is more vulnerable to the influence of public opinion,and the evolution of public opinion is more polarized;The emotional fluctuation in central big cities is relatively stable,and the negative emotions of netizens in other regions are often caused not by the epidemic itself,but by the negative public opinion caused by the epidemic;Compared with ordinary netizens,influential groups have more serious negative emotions in the stage of public opinion evolution.Emotional evolution presents obvious ripple effect in each stage.It is necessary to formulate effective public opinion guidance policies for different groups in different stages.【Innovation/limitation】This paper divides the whole corpus into more than 50 small corpora.Very few small corpora have too little text information,which has certain limitations.