With the development of mobile Internet, Weibo has accelerated the spread of social emergencies. Retweeting as the most important users' information behavior guide the trend of network public opinion to a great extent. Therefore, this study takes Hangzhou nanny arson as an example of social emergency, combining psychology process words to construct a VAR (Vector autoregressive) regress model. According to the results of empirical study, social and affect process words in LIWC can have a predictable role for retweeting behavior. This study expands the research field of psycholinguistics in the context of social emergencies, furthermore, it also provides inspiration for the government or enterprise emergency management departments to manage public opinion on the network.