Gugak Analysis and Education

Main Contributors: Danbinaerin Han

Objective: Computational Analysis and Educational Research on Korean Traditional Music (Gugak)

Background

Gugak is a broad term encompassing traditional Korean music and its surrounding culture, including court music, chamber music (pungnyu), and art music that have been transmitted across generations on the Korean peninsula. Gugak is rooted in the systematic combination of melody and expression, and although indigenous notation systems such as jeongganbo exist, actual transmission has primarily relied on oral tradition and master-to-student apprenticeship (gujeonsimsu). These transmission practices and musical structures call for modeling strategies that differ fundamentally from those developed for Western music, which has evolved around tonal-harmonic systems and standardized notation traditions.

While the Music Information Retrieval (MIR) community has grown rapidly over the past two decades, its research has largely developed under the assumptions of Western music theory—where tonal systems, metric structures, pitch frameworks, and terminological definitions are relatively well-agreed upon. In other words, conventional MIR has been built on the premise of stable annotation schemes and relatively clear ground truth [1].

In the case of Gugak, however, the interpretation and categorization of core concepts such as jangdan (rhythmic cycles), eumjojik (tonal organization), and sigimsae (ornamental techniques) have been debated within the scholarly community for a long time and have yet to converge on a single set of definitions. These disagreements are not merely terminological; they reflect deeper differences in how musical structure and expression are understood.

Consequently, the process of computationally modeling Gugak—reducing performance traditions into discrete labels and quantifiable representations—inevitably involves choices about how to define categories and draw boundaries. Nevertheless, the need to computationally address Gugak continues to grow across diverse areas including digital archiving, music education, automatic analysis, and generative model research. These demands call for a new direction in MIR research that considers social context and educational applicability from the very stage of dataset design.

Topic 1: Gugak Datasets

In this context, building Gugak datasets is understood as a translation process—redefining complex musicological concepts into computationally tractable formats. Rather than presupposing a single correct answer, this work is closer to establishing agreeable standards amid multiple possible interpretations, while explicitly documenting the choices made along the way.

We are currently building Gugak datasets in collaboration with the National Information Society Agency (NIA), the National Gugak Center, and the Korea Culture Information Service Agency. The dataset is designed to include not only traditional musical metadata such as genre, constituent tones, and jangdan, but also performance-based expressive information such as mood, vocal technique (changbeop), and sigimsae. In particular, by combining expert consultation with crowdsourcing, we are constructing metadata that reflects multiple layers of perception rather than a single authoritative interpretation.

In addition, we are digitizing and publicly releasing symbolic data of jeongak (Korean court and literati music) based on jeongganbo notation through Optical Music Recognition (OMR). This effort aims to extend traditional notation systems for use in modern computational environments.

Jeongak dataset download: https://www.danbinaerin.com/Jeongganbo_dataset/

Topic 2: Structural Analysis of Korean Folk Songs

Tosok minyo (indigenous folk songs) are music formed and transmitted within the everyday lives of ordinary people rather than by professional musician communities. Closely tied to social contexts such as labor, play, and ritual, these songs exhibit strong regional variation and variability, making them an important subject of research in terms of scale structure and musical organization.

Because orally transmitted folk songs involve flexible combinations of melody, lyrics, and breathing units that adapt to repetitive variation and performance context, it is difficult to directly apply the fixed segmentation units assumed in Western music structural analysis. In particular, the challenge of objectively defining motif boundaries has been identified as a core difficulty in folk music MIR.

In this research, we focus on the close relationship between lyric-based units, semantic segmentation, and musical motif formation. We manually annotated lyric-based motif boundaries and used these annotations to fine-tune an automatic speech recognition model. This enabled automatic motif segmentation across a large-scale folk song dataset (Han et al., 2025).

Our analysis revealed that motif density and length variability (entropy) show statistically significant differences according to the social function of the songs—such as labor songs, play songs, and lullabies. This quantitatively demonstrates that the musical structure of indigenous folk songs is not merely a matter of acoustic patterns but is closely tied to social function and performance context. This research presents a scalable computational method for structural analysis of oral tradition music and suggests the potential of context-aware MIR approaches for non-Western music.

Related Publications

• Motive-level Analysis of Form-functions Association in Korean Folk Song
  Danbinaerin Han, Dasaem Jeong, Juhan Nam
  arXiv preprint, 2025
  [paper]
• Finding Tori: Self-supervised Learning for Analyzing Korean Folk Song
  Danbinaerin Han, Rafael Caro Repetto, Dasaem Jeong
  Proceedings of the 24th International Society for Music Information Retrieval Conference (ISMIR), 2023
  [paper]

---

References

• [1] Gómez-Cañón, Juan Sebastián, et al. "Beyond a Western Center of Music Information Retrieval: A Bibliometric Analysis of the First 25 Years of ISMIR Authorship." Transactions of the International Society for Music Information Retrieval 8.1 (2025): 372-387.