Figure 1
Overview of the research framework.
Figure 2
Measure‑level alignment between eight different versions of Yeominlak, in chronological order (the earliest is at the top), provided by NGC.
Figure 3
An example output generated by a model trained using the Yeominlak chronological score (Chwipunghyeong).
Figure 4
An example orchestration output generated by a model trained using the contemporary Yeominlak score (Chwipunghyeong, second measure).
Figure 5
An example of Jeongganbo in the original notion (below) and a broadly equivalent conversion to Western classical notion (above). Dashed lines are not part of either notation and are added simply to clarify the temporal alignment between the two systems.
Figure 6
Jeonggan‑like encoding position labels.
Table 1
Ensemble configurations of the Jeongganbo dataset.
| Ens. Size | Pieces | Instrumentation |
|---|---|---|
| 6 | 18 | Daegeum, Piri, Haegeum, Ajaeng, Gayageum, Geomungo |
| 5 | 54 | Daegeum, Piri, Haegeum, Gayageum, Geomungo |
| 4 | 19 | Daegeum, Piri, Haegeum, Ajaeng |
Table 2
Cycle length distribution and token variety across instruments.
(a) Total number of jangdan (daegeum)
| Cycle Length | Pieces by Ensemble | Total No. of Jangdan | ||
|---|---|---|---|---|
| 4‑Ens. | 5‑Ens. | 6‑Ens. | ||
| 20 | 3 | 2 | 2 | 219 |
| 18 | 2 | — | — | 30 |
| 16 | 4 | 31 | 5 | 543 |
| 12 | 5 | 8 | 3 | 569 |
| 10 | 2 | 7 | 5 | 336 |
| 8 | 1 | 1 | — | 71 |
| 6 | 2 | 4 | 3 | 429 |
| 4 | — | 1 | — | 67 |
| Other | — | — | — | 101 |
(b) Unique and total token types per instrument
| Instrument | Unique Tokens | Total Tokens | ||
|---|---|---|---|---|
| Pitch | Sigimsae | Pitch | Sigimsae | |
| Daegeum | 21 | 44 | 28516 | 19864 |
| Piri | 21 | 40 | 25869 | 15452 |
| Haegeum | 18 | 31 | 24811 | 9106 |
| Gayageum | 19 | 2 | 20722 | 39 |
| Geomungo | 19 | 19 | 18837 | 4082 |
| Ajaeng | 15 | 7 | 8352 | 623 |
Figure 7
Comparison between Jeonggan (JG)‑like, REMI‑like, and ABC‑like encoding schemes.
Table 3
Average token counts per instrument per piece with jeonggan‑like encoding (18 pieces with 6 ens).
| Instrument | Total | Pitch | Sigimsae |
|---|---|---|---|
| Daegeum | 1876.9 | 470.6 | 343.9 |
| Piri | 1658.1 | 428.7 | 258.3 |
| Haegeum | 1388.9 | 395.0 | 118.1 |
| Gayageum | 1086.1 | 346.3 | 0.9 |
| Geomungo | 1119.2 | 326.8 | 66.7 |
| Ajaeng | 1042.3 | 306.5 | 22.5 |
Figure 8
Orchestral part generation.
Table 4
Quantitative evaluation results. “Geom.” denotes geomungo, “Daeg.” denotes daegeum.
| Piri to Geom. | Every to Daeg. | |||
|---|---|---|---|---|
| len‑mat | F1 | len‑mat | F1 | |
| JG‑like | 0.978 | 0.638 | 0.992 | 0.644 |
| REMI‑like | 0.997 | 0.624 | 0.992 | 0.580 |
| ABC‑like | 0.904 | 0.578 | 0.930 | 0.571 |
| JG with subgenre | 1.0 | 0.682 | 1.0 | 0.694 |
Figure 9
Generated output of the daegeum part of Chwipunghyeong.
Figure 10
Screenshot of the interactive web application.
Figure 11
Performance in commemoration of King Sejong’s birth anniversary. Photo courtesy of the NGC.
Figure 12
Technology and performance showcase. Photo courtesy of the NGC.
Figure 13
Open discussion session. Photo courtesy of the NGC.