Scholarlink

Depression is a critical global health challenge, with news media playing a vital role in shaping public understanding. This study employs computational methods to analyze depression news coverage in China and the U.S., examining 542 articles from People’s Daily Online and Cable News Network (CNN) (2022–2024). We develop a Large Language Model (LLM)-assisted framework integrating Health Belief Model (HBM) content coding and Latent Dirichlet Allocation (LDA) topic modeling. Our findings reveal significant cross-cultural differences: People’s Daily Online emphasizes risk information and youth-focused interventions through family and school systems, while CNN highlights individual experiences, self-empowerment, and practical advice. The LLM-based coding achieved 0.68–0.82 inter-coder reliability coefficients across HBM constructs. Co-occurrence analysis shows People’s Daily Online’s sparse construct networks creating risk-warning narratives versus CNN’s dense interconnections building mobilization pathways. These discoveries illuminate divergent health communication strategies and provide empirical evidence for optimizing depression reporting across cultural contexts.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

LLM-Assisted Text Mining Framework for Cross-Cultural Health Communications: Analyzing Depression News Coverage in Chinese and American Media

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ruijia Zhang,
  • Chaoyuan Zuo

摘要

Depression is a critical global health challenge, with news media playing a vital role in shaping public understanding. This study employs computational methods to analyze depression news coverage in China and the U.S., examining 542 articles from People’s Daily Online and Cable News Network (CNN) (2022–2024). We develop a Large Language Model (LLM)-assisted framework integrating Health Belief Model (HBM) content coding and Latent Dirichlet Allocation (LDA) topic modeling. Our findings reveal significant cross-cultural differences: People’s Daily Online emphasizes risk information and youth-focused interventions through family and school systems, while CNN highlights individual experiences, self-empowerment, and practical advice. The LLM-based coding achieved 0.68–0.82 inter-coder reliability coefficients across HBM constructs. Co-occurrence analysis shows People’s Daily Online’s sparse construct networks creating risk-warning narratives versus CNN’s dense interconnections building mobilization pathways. These discoveries illuminate divergent health communication strategies and provide empirical evidence for optimizing depression reporting across cultural contexts.