Framing the COVID-19 Crisis: Analyzing State Media Strategies in China Using Deep Learning Models

This study examines China's crisis communication strategies during the COVID-19 pandemic through computational analysis of state media. Using Convolutional Neural Networks (CNN) and logic-tree analysis, we analyzed 17,631 sentences from Xinwen Lianbo broadcasts (January–September 2020) to identify framing patterns in centralized governance systems. Our analysis revealed 16 distinct frames organized hierarchically, encompassing both substantive (policy measures, epidemic information) and symbolic (political narratives, unity themes) dimensions. Three key findings emerged: (1) China maintained strategic equilibrium between substantive and symbolic framing to balance operational competence with ideological legitimacy; (2) communication strategies demonstrated dynamic phase-based adaptation while preserving core political narratives; (3) innovation occurred primarily in technical domains rather than fundamental communication approaches. These findings extend crisis communication theory beyond Western democratic contexts, revealing how centralized systems achieve adaptive flexibility through coordinated channels while maintaining message consistency.