A Multimodal Knowledge Management Approach to Korean Classical Literature in Digital Communication

Korean classical literature embodies rich but fragmented cultural knowledge—such as rituals, poetic imagery, and recitation practices—that risks loss under text-only digitization. To address this, the authors propose a multimodal knowledge management framework that captures, aligns, and reuses cultural knowledge from sources like Samguk Yusa and the Annals of the Joseon Dynasty by integrating text, historical illustrations, and reconstructed audio. They build a fine-grained corpus using layout-aware OCR and phoneme alignment, then develop a cyclic attention fusion (CAF) model with mutual information constraints to ensure cross-modal semantic coherence. Evaluation combines BLEU-4/TER with new cultural metrics—Cultural Adequacy Score (CAS) and Cross-Semantic Alignment Index (CSAI)—showing improved fidelity in representing customs, narrative flow, and ritual context. The work contributes to knowledge management by treating intangible heritage as structured, multimodal knowledge assets, and offering a reusable AI-driven system for cross-cultural knowledge reuse and public knowledge services.