An Embodied Cultural Heritage Semantic Mining Approach Driven by Bidirectional Collaborative Mechanism

ABSTRACT Embodied cultural heritage information resources carry the bodily experiences and perceptual logic formed through human cultural practices, exhibiting high multimodality, immersiveness, and dynamic transmission characteristics. However, traditional methods of information organization struggle to reveal the tacit embodied knowledge within them. To address this, this study focuses on typical embodied cultural heritage information resources represented by the guqin jianzi notations, constructing an embodied semantic ontology model and proposing a bidirectional collaborative mechanism integrating large language model with ontology model to enhance the semantic understanding of embodied knowledge by large language models. The study constructed two embodied knowledge base, puzi knowledge base and jianzi knowledge base, and conducted multiple experiments to evaluate the performance of large language model in understanding the semantics of single jianzi and jianzi sequences. The results show that the embodied knowledge bases significantly enhance large language model's ability to interpret complex cultural symbols, particularly improving accuracy in understanding embodied semantic dimensions such as fingering techniques, hui positions, and timbre. Theoretically, this study expands new pathways for semantic modeling of embodied cultural heritage; practically, it provides a feasible technical solution for the digital preservation and intelligent transmission of embodied cultural heritage.