Sentiment-topic dynamic collaborative analysis-based public opinion mapping in aviation disaster management: A case study of the MU5735 air crash

Abstract

In the new media era, public opinion guidance for major emergencies is significant for enhancing social governance capabilities and ensuring social stability. Constructing a public opinion map based on sentiment-topic dynamic collaborative analysis can serve as a valuable reference for government-led public opinion guidance. This paper firstly develops a sentiment dictionary for major disaster events to aid in recognizing netizen emotions and uses a relevance formula-based Latent Dirichlet Allocation (LDA) model to conduct topic text mining. Then, through the division of public opinion cycles and different methods-based topic classification, sentiment-topic dynamic collaborative analysis is performed to create a visual public opinion map. By examining the case of the China Eastern Airlines MU5735 crash, the results indicate a close relationship between netizens’ sentiments and the topic content, primarily influenced by mainstream media news reports during different stages of public opinion. Furthermore, it is find that certain key node events can trigger emotional resonance among the masses during different stages of the public opinion cycle, resulting in completely different sentiments such as “sorrow”, “good” and “anger” for different topics. The visual public opinion map generated in this study offers valuable insights for the public opinion guidance and supervision of related disaster events. Simultaneously, it can also help identify and understand potential public opinion risks of disaster events, such as the spread of rumors and the amplification of panic. This provides both researchers and government administrators with a comprehensive and intuitive tool to better understand public opinion and effectively mitigate the public opinion risks associated with disaster events, thus safeguarding public safety and interests.