This study proposes a computational analysis-oriented framework for analyzing how digital media art can enhance corporate governance information expression and influence investor behavior. A multi-modal data set containing 3240 corporate governance disclosure samples and 25600 investor interaction records was constructed, covering the governance chapter of the annual report, the ESG governance page, the investor relations webpage and the governance theme visualization page. A structured coding module is designed to convert visual rhetoric and governance semantics into a unified feature tensor, and a cross-modal fusion analysis module for investor behavior recognition is constructed to incorporate governance text nodes, visual structure nodes and interaction logs into a unified discriminant space. The experimental results show that the proposed method achieves 87.41% classification accuracy, 0.861 F1 value, 0.912 AUC and 0.793 MCC. The results show that the expression of corporate governance information empowered by digital media art can be effectively transformed into computable behavior representation, and provide support for the analysis of the expression effect of governance information and the identification of investor behavior.
