Research on the Evolutionary Path of Public Opinion in Networked Emergencies Based on Large Models——Taking the Gansu earthquake as an example

Construct a rational map of online public opinion on emergencies, reveals the evolutionary paths and developmental contexts of public opinion incidents, providing decision support for government departments in managing and guiding public sentiments. Select Sina Weibo as the data source, use octopus collector to collect Gansu earthquake Weibo data, and then perform data preprocessing. The causal relationships of the events were determined based on a rule-based template matching method, and a multi-round dialogue mechanism based on large language models was used to automatically extract causal event pairs. The K-means++ clustering algorithm was employed to generalize the events and construct a network public opinion event logic graph, analyzing the evolutionary paths of online public sentiments. The research findings indicate that public sentiment regarding the Gansu earthquake is characterized by multiple levels and complex causal relationships. The involvement of multiple causal chains contributes to a high degree of complexity in the evolution of public sentiment. By constructing an event logic graph, the development process of the Gansu earthquake public opinion event has been demonstrated, which can provide technical support for real-time monitoring, in-depth analysis, and precise response of public opinion.