| In 2021,against the backdrop of the Delta variant that had been rampant around the world,small-scale outbreaks triggered by it in China had occasionally broken out and had been quickly contained.During such public health emergencies,Weibo,short videos and other social media platforms,which are characterized by fast dissemination and high public opinion heat,are important platforms for public opinion generation and dissemination,and timely control and guidance of such platforms can avoid the consequences of triggering social unrest and loss of government credibility.Some studies have shown that the differences and evolutionary trends of public opinion are correlated with geographical distribution,but most of the current studies on public health emergencies focus on the evolution trend of topics and sentiments over time,and there is a lack of studies that consider spatial factors.Therefore,this paper selected public health emergencies with different locations and certain time intervals as research cases,took the Weibo texts published during the Delta variant outbreak as research objects,adopted topic mining model,quantitative sentiment analysis,sentiment map and other research methods,built a topic model and a sentiment model based on topic mining and sentiment cognition theory,combined with temporal and spatial factors.This model was constructed to visually analyze the topic and sentiment of public health emergencies in different time and space,and to provide suggestions and decisionmaking basis for public opinion guidance.This paper firstly proposes an analysis scheme combining qualitative research and quantitative analysis,then constructs a topic model and a sentiment model,and finally verifies the validity of the model through specific events,and uses graphic visualization techniques to express the quantitative results.This study consists of three major parts:data acquisition,data modeling,and data analysis.After crawling the original Weibo corpus using crawlers to extract keywords,the corpus is selected by key words,and the original data is processed by de-duplication and word separation;the cleaned corpus is processed by the improved Jieba lib,and the Snow NLP lib is optimized with another word separation method.The LDA model is determined by the number of topics at different phases,meanwhile the sentiment value is calculated using the Snow NLP lib,and the sentiment map within the life cycle of the outbreak is fitted using Gephi software;combined with specific cities and periods,the evolution direction of Weibo public opinion over time and space could be analyzed according to the topic-word heat mined by LDA model and the sentiment value calculated by Snow NLP lib,the evolution law could be found out,and the pervasive development order of unexpected public health emergencies could be summarized.This paper takes public health emergencies in Shijiazhuang,Ruili and Nanjing as research objects,and reveals the spatial and temporal characteristics and sentiment evolution of online public opinion topics through topic mining,quantitative sentiment analysis and sentiment map.The study found that during the epidemic period,there is a significant concentration of public opinion topic heat on the topics of epidemic notification and epidemic prevention and control,the topic heat slightly decreases during the epidemic receding period;the overall public opinion heat is higher in cities with high public participation,open culture and more developed economy,and online public opinion topics are more dispersed;special time nodes,such as the Spring Festival and other holidays,also have a certain influence on the heat of a specific public opinion topic;in terms of sentiment evolution,all three cities show a trend of first decreasing and then increasing positive sentiment,Shijiazhuang and Nanjing are always in neutral to positive sentiment,and Ruili has a higher percentage of negative sentiment,which is due to the level of material supply,economic status and the optimism of the public about the prevention and control of the epidemic,Shijiazhuang has a greater percentage of negative sentiment than Nanjing.Due to the long-term border control pressure,the agile response to trace and control the epidemic in Ruili takes longer than the economically abundant cities,which makes the sentiment presented on social platforms such as Weibo will be more negative than inland cities,while also reflecting the severity of the epidemic to some extent.In terms of monitoring and management,cities with large volumes and high heat of public opinion should monitor the trends promptly to avoid causing public opinion crisis,and cities with large fluctuation of emotion and a high percentage of negative emotion should realize the establishment of early warning system and timely guidance,and establish initiatives to deal with negative emotion as soon as possible.This study can provide a theoretical and practical basis for feedback on online public opinion,prior planning of proposals,grasping public opinion trends,and governing online public opinion. |
PDF Full Text Request