【Purpose/significance】 A series of hot events during the epidemic in Xi'an brought many negative comments, making it par⁃

ticularly important for the government to improve the public opinion guidance strategy in the context of normalized epidemic preven⁃tion and control. This paper explores the ability and extent of the influence of official informal tweets on public sentiment in the Inter⁃net environment, which helps to improve the effectiveness of public health emergency information disclosure and enable information to empower the practice of epidemic prevention and control.【Method/process】 Conduct topic clustering and demand extraction for micro⁃blog user comments, and code the official blog post type, and measure the relationship between the type of social platform information release and the public emotion with the public demand as the intermediary.【Result/conclusion】 The research uses K-means clustercoding to get 10 topics and 47 sub-focuses, and draws a demand map. Through Spearman correlation coefficient, it finds that the epi⁃demic situation broadcast types of five blog posts have different effects on emotions. Based on the research findings, it proposes a tar⁃geted government strategy response model.【Innovation/limitation】This article innovatively adopts the "big data+small data" research method to explore the correlation between user emotions and government information release in emergency situations from a micro per⁃spective. We look forward to further exploring the impact and causality of "Govinfo-emotion" from multi-source heterogeneous data in the future.