In the age of generative AI, the proliferation of Western-biased music models at home and abroad necessitates a rapid response from the Gugak (Korean traditional music) community. Accordingly, this study explored methods for transforming existing Gugak archives into AI training data. In the field of Gugak, structured metadata has been central to data description (記述), both in archiving and in the construction of training data. Conversely, data description within the AI ecosystem is increasingly trending towards unstructured natural language narratives. Amid this trend, recent research has demonstrated that hybrid AI models—which combine structured data with descriptive, natural language-oriented data—improve both the generation quality of the output and user controllability. Therefore, this paper proposes advancing toward a hybrid data curation model by upcycling existing domestic archive data, such as that held by the National Gugak Center.
Building similarity graph...
Analyzing shared references across papers
Loading...
Byung-O Kim
The humanities of coexistence
Building similarity graph...
Analyzing shared references across papers
Loading...
Byung-O Kim (Sat,) studied this question.
www.synapsesocial.com/papers/69abc0de5af8044f7a4e97dd — DOI: https://doi.org/10.37524/huco.2026.01.22.125