Abstract: The spread of public opinion caused by the outbreak of public emergencies has caused major distress to all sectors of society. An effective method is to find public opinion points in advance and stifle public opinion in the cradle. A system including visual analysis results and attribution analysis is constructed. The system takes public opinion events as the theme, uses data collection technology, and combines sentiment analysis and text processing methods to monitor and analyze public opinion information. Experiment results show that this system is effective for the analysis of public emergencies. By monitoring the changes in mood fluctuations in the process of public opinion, the purpose of assisting the government in decision-making is finally achieved, laying a foundation for the analysis of future public emergencies.

Key words: public emergencies, public opinion analysis, emotional changes, visualization