Abstract: 【Purpose/significance】With respect to public health emergencies, event evolutionary graphs are constructed on different platforms. By integrating sentiment analysis with the evolution of geographical influence, this approach provides an evolutionary analy⁃ sis.【Method/process】Utilizing data from Weibo and Bilibili concerning "Mycoplasma pneumonia," our approach involves the applica⁃ tion of rule-based templates to distill and discern the causal, sequential, parallel, conditional, and reversal relationships inherent in the event narratives. Leveraging the BERTopic model for event generalization, we construct an abstract event evolution graph, reveal⁃ ing the patterns of public opinion evolution from multiple perspectives of sentiment and geographical influence.【Result/conclusion】 Our findings reveal that network public opinion events exhibit intricate interconnections in multiple dimensions, showing substantial variations across diverse platforms. The logical characteristics of the event evolutionary graph can analyze the evolution path of online public opinion from multiple dimensions, which can help public opinion control departments formulate more effective governance strat⁃ egies.【Innovation/limitation】Event generalization using BERTopic to construct a matter-of-fact map based on multiple logical rela⁃ tionships to analyze the evolution path of online public opinion. Given the limitation of the time span of the text, it is necessary to ex⁃ pand the time range subsequently.