Evolution of online public opinions on major accidents: Implications for post-accident response based on social media network

Based on the incredible breadth and speed of information distribution within social media network, and continuous development of natural language processing techniques for sentiment and topic analyses, it is potential to access online public opinions on major accidents. An integrated framework including four-stage evolution model, lexicon-based sentiment analysis, and latent Dirichlet allocation (LDA)-based topic extraction was developed based upon social media data analysis. The explosion of Xiangshui eco-chemical industrial zone (EXEIZ) was taken as a case for investigating the evolution of online public opinions. The large accident domain (LAD)-based sentiment dictionary was established for the lexicon-based approach for sentiment determination. Effective methods (i.e., combination of short microblogs and comments into a new text, and the hybrid approach integrating perplexity with principal component analysis) were deployed for overcoming two typical shortcomings (i.e., inappropriate for short text dataset, and sensitive to the number of topics within a corpus) of the LDA-based approach for topic analysis and extraction. According to the four-stage evolution model, customized strategies and suggestions were provided for guiding and controlling online public opinions on major accidents, in order to decrease their negative impacts on the society. This enabled conducting targeted communication efforts among different stakeholders for the reduction of negative sentiments such as indignation, sadness, and threat, and the avoidance of social media crises. Although a major accident poses tangible threats to the public, it may improve their awareness for preventing from these threats. In case of appropriate measures in time, the focus tends to steer toward effective and prosocial solutions. It is helpful for sustaining and re-establishing the image of authorities, enterprises or individuals that are closely associated with the major accident.