Abstract:With the development of internet, the net becomes an increasing important channel for people to post, access, and exchange information. By 2012, the number of global internet users had reached 2.5 billion. The number of new users in Asian countries accounted for 54%, much higher than that in other continents. Among these countries, the Chinese users rose to 0.56 billion with an average time of accessing internet 2.9 hours per day. However, due to lack of authority and authenticity, the internet media gradually becomes a breeding ground of public opinion crises. For example, the 2010 Gulf of Mexico oil spill, the 2011 China Wenzhou train collision, the 2012 Islands dispute in East Asia, the 2013 PRISM incident, and the 2014 Malaysia Airlines plane crash caused great attention and wide discussion on the internet, which further developed into unpredictable online public opinion crises. It brings many difficulties to the response of online public opinion crises. The phenomenon has drawn increasing attention from social scientists around the world. Therefore, the paper focuses on the intervention mechanism of online public opinion crisis. After collecting and sorting several public opinion crisis incidents in recent years, the basic expression forms, driving factors, and specific patterns of public opinion spread are systematically summarized. Then considering the general dissemination and spread process of online public opinion. The intervention mechanism can be divided into four parts in this paper. The first part is the judgment for public opinion intervention. By calculating and comparing the relevant indexes, we will further determine if the public opinion crisis incidents need to be intervened in this part. The second part is the time selection for public opinion intervention. Considering the high similarity of public opinion spread process of similar events, the paper summarize the general dissemination process of 8 classes of public opinion crisis events. Through analyzing the variation regularity of similar events, we could find the best time for intervention in this part. The third part is the intervention strategy choice. After evaluating the level and type of different social events, we will choose the better set of intervention strategies in this part. The specific intervention strategies include isolation strategy, insertion strategy, and reconstitution strategy etc. The last part is the effect evaluation for intervention. We will use some evaluation methods and strategies, such as entropy evaluation method, expert evaluating method, and fuzzy comprehensive evaluation, to assess the implementation effect of different intervention mechanisms in this part. Considering the principles of usefulness, comprehensiveness, scientificalness, and applicability, this study also discusses the technical solutions and effective ways of each part of intervention mechanism. The related core technologies mainly include public opinion monitoring system, propagation process simulation, network evolution simulation, and effect evaluation for intervention. Public opinion monitoring system is mainly used in data acquisition of different social events. After collecting the online data in the field of occurrence time, incident location and event attribute, propagation process simulation can be used to determine the diffusion process of public opinion. Network evolution simulation is the tool to analyze the possible evolution pattern of public opinions. Besides, this paper also gives policy proposals for the intervention of public opinion, which could provide references for the government or relevant departments in the disposal of online public opinion crises. The specific policy proposals are shown as follows. The first proposal is to establish the legal system for the management of network public opinion. The second proposal is to classify the public opinion and develop the differentiated strategies for social events. The third proposal is to establish a long-term monitoring and early warning mechanism of online public opinion.