notesum.ai
Published at November 27Fusion of Discrete Representations and Self-Augmented Representations for Multilingual Automatic Speech Recognition
cs.SD
eess.AS
Released Date: November 27, 2024
Authors: Shih-heng Wang1, Jiatong Shi2, Chien-yu Huang1, Shinji Watanabe2, Hung-yi Lee1
Aff.: 1National Taiwan University; 2Carnegie Mellon University

| LS | MS | Bitrate | ||||
|
- | 2.34 | 10.89 | 2048000 | ||
|
- | 2.37 | 22.4 | 356.19 | ||
| MMS-1B | - | 2.32 | 14.32 | 280.86 | ||
|
- | 2.52 | 14.38 | 556.15 | ||
| Concat MMS-1B |
|
1.92 | 11.25 | 665.13 | ||
| MMS-1B |
|
1.89 | 10.87 | 665.13 | ||
| MMS-1B |
|
2.26 | 12.22 | 1024.90 | ||
| MMS-1B |
|
2.17 | 11.69 | 648.52 |