Share
Export Citation
Expert-Anchored Semantic Cross-Modal Dance Clustering for Sustainable Intangible Cultural Heritage Preservation
Erna A.
Proceedings 2026 3rd International Conference on Technology Innovation and Its Applications Ictiia 2026
Abstract
Intelligent systems capable of capturing the meaning of movements and preserving cultural interpretations are essential for the sustainability of traditional dance. This study introduces an expert-anchored, knowledge-guided cross-modal clustering framework that integrates pose data (reduced via Principal Component Analysis) with IndoBERT text features (reduced via Linear Discriminant Analysis), weighted by alpha. Semantic anchors are constructed based on expert narratives and correlated with pose samples according to their respective class labels. These labels are directly used by the LDA to learn the discriminative semantic representations and guide the clustering stage. A full-covariance Gaussian Mixture Model with Hungarian realignment ensures semantic consistency. Experiments on 144 Paduppa dance videos show that expert anchors reduce multimodal differences, strengthen cultural understanding, and support reliable digital heritage management, providing a scalable foundation for ethical, human-centered AI in cultural preservation.