Exploring Chinese Youths’ Focus and Sentimental Tendencies Toward Different Extreme Weather Events: A BERTopic and Zero-Shot GPT-4o-mini Analysis

Extreme weather events increasingly intensify vulnerabilities in social, economic, and environmental systems. Typhoons damage infrastructure and disrupt mobility, cold waves threaten public health and energy demand, heatwaves strain electricity grids and exacerbate health risks, while rainstorms often cause urban flooding and damage crops. Understanding how the public sentimentally and cognitively responds to these events is essential for improving climate communication and resilience strategies. This study adds insights by applying a deep learning-based multi-model framework on Bilibili, a major Chinese youth-oriented social media platform. Novel in both method and context, the study integrates BERTopic for topic modeling and a zero-shot, prompt-based GPT-4o-mini model for sentiment classification -- representing a rare application of this combined approach in Chinese climate discourse. It also contributes a comparative perspective by examining youth reactions to four distinct extreme weather types. We analyzed 443,157 user comments on 182 videos posted by the nongovernmental organization (NGO) account “Chinese Weather Enthusiasts” throughout 2023. Results revealed dominant negative sentiment across all weather types, with rainstorms (51.90%) and heatwaves (49.63%) generating the strongest reactions. Typhoon discussions focused on emergency preparedness and regional comparisons; cold waves evoked health concerns and nostalgia; heatwaves prompted frustration over physical discomfort; and rainstorms drew criticism of infrastructure and public safety. These findings reveal sentimental and topical responses among Chinese youth. By leveraging social media data, the study provides actionable insights for policymakers, NGOs, and emergency planners to develop more targeted communication, improve infrastructure, and strengthen societal resilience to climate-related crises.