Abstract: 【Purpose/significance】In the era of Web3.0, the network public opinion of public emergencies is deeply embedded in the field of social risk governance. Accurately assessing the risk of public opinion on the Internet of public emergencies is the premise and basis for effective early warning and response to risk challenges【. Method/process】Based on the tipping point theory, this study con⁃ structs a multimodal public opinion risk assessment system under the three-dimensional coupling framework of "adhesion - individual characters - Environmental Power". The entropy weight TOPSIS method is used to quantify the public opinion risk indicators, and the K-means clustering algorithm is combined to achieve the objective classification of risk levels, so as to identify low, medium and highrisk public opinion events【. Result/conclusion】The study found that the risk level of online public opinion of public emergencies was affected by multiple factors, such as event content orientation, opinion leaders' communication power, algorithm boost, government governance and the degree of overseas intervention, among which the participant dimension and environment dimension were the key drivers of public opinion risk evolution. The research proposes a comprehensive governance framework of "1 system+2 mechanism+n" for high-risk public opinion, which provides a solution with both theoretical explanatory power and practical operability for dealing with network public opinion of public emergencies in the Web3.0 era.【Innovation/limitation】The creative introduction of "overseas participation" index has formed an evaluation system covering domestic and overseas double cycle risk factors. However, the sample database is relatively single, so we should consider expanding the case database in the future to enhance its richness and integrity.