In this paper, we collected more than 60,000 weibo comments data from 2020 January 20 to December 28, by Python crawler. Subsequently, we used the SnowNLP model based on the naive Bayes algorithm to classify the text corpus with sentiment orientation, analyzed the evolution of epidemic-related topics, and visualized the display from the two dimensions of time and space. On temporal dimension, the emotional attitudes of netizens experienced an anxiety fluctuation period, a stable transition period, and a period of deterioration in public opinion during the beginning of the outbreak (January 20-April 28). The overall emotional attitude of netizens showed negative characteristics. netizens' sentiment experienced a period of rising volatility and steady improvement, with positive sentiment dominating during the normalization phase (May 1st-December 28th). On spatial dimension, we found that there were significant differences in the emotional state and attention of users in different administrative regions with geographic statistical analysis. The more severe the epidemic situation, the higher the topic participation of weibo users and the lower the emotional index. This research provides theoretical reference and event significance for targeted public opinion guidance at the macro level.
Wan, Jiangping; Liu, Xu; Zuo, Yihang; and Luo, Jianfeng, "Analysis on Public Opinion Sentiment Evolution of COVID-19 Based on Weibo Data" (2021). WHICEB 2021 Proceedings. 67.