Weibo public opinion analysis of emergencies using lexicon-based and deep learning approaches: a case study of the 12.18 Jishishan Seismic Event

Seismic events, as sudden natural disasters, significantly impact society and the economy. Analyzing post-disaster online public opinion helps quickly assess the situation, severity, and public needs, aiding sentiment management and emergency response. This study collected Sina Weibo public opinion data within 24 h after the Ms6.2 earthquake that struck Jishishan County, Gansu Province, on December 18, 2023. The distribution characteristics of public opinion and micro-charity and the correlation between public attention and micro-charity participation were analyzed. Sentiment analysis was further conducted using an improved sentiment lexicon-based approach and the Text-CNN method. The results indicate that the public reactions were most intense within the first hour after the earthquake, followed by three subsequent fluctuations. In terms of sentiment, positive-sentiment blog posts were the most common. Regarding micro-charity, users in regions highly concerned about the ‘#Gansu Earthquake’ topic were more involved. Different from previous studies, this study reveals a positive correlation between media exposure and micro-charity, suggesting that provinces with greater emergency attention are more active in micro-charity. The findings of this study are significant for disaster emergency management and online public opinion analysis in the new era.