Research on optimizing the heritage transmission path of Muyu Song using AI technology

To address challenges in digital dissemination of Muyu Song intangible cultural heritage—including inefficient audience matching, rigid dissemination path scheduling, and delayed effectiveness feedback—this study models the dissemination process as a sequential decision problem. It constructs a computational framework integrating multimodal feature representation, dissemination effectiveness prediction, and path optimization. Centering on the collaborative modeling of audio rhythmic features and lyrical semantic information, a unified segment-level multimodal representation is formed. Building upon this, a dissemination effectiveness prediction mechanism is introduced to jointly estimate key metrics including clicks, retention, interaction, and diffusion. Further integrating Markov decision processes with reinforcement learning strategy updates enables dynamic adjustment of dissemination actions across platforms and time scales. On held-out songs, the five prediction heads achieved R 2 values of [range], while component ablations consistently confirmed that rhythm-aware segmentation, dialect-preserving encoding, and cultural constraints improved retention, interaction, diffusion depth, and traditional-context fidelity over the generic multimodal baseline. The relevant research results indicate that integrating the dissemination of intangible cultural heritage into a calculable decision-making and optimization system can help reveal the intrinsic connections among content characteristics, audience reactions, and platform mechanisms, providing a verifiable analytical framework for the construction and intelligent digital dissemination of intangible cultural heritage.