Automatic Speech Recognition on Swivuriso and NCHLT Combined Average
28.41WERW2V-BERT Tone-cond.
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| W2V-BERT Tone-cond.Protocol=Fine-tuned, Architecture=W2V-BERT, Strategy=Tone-conditioned2026.06 | 28.41 | 6.98 | 93.4 | |
| W2V-BERT Tone+Curr.Protocol=Fine-tuned, Architecture=W2V-BERT, Strategy=Tone-conditioned + Curriculum2026.06 | 29.38 | 7.25 | 93.1 | |
| Whisper MultilingualProtocol=Fine-tuned, Architecture=Whisper, Strategy=Multilingual Baseline2026.06 | 29.44 | 7.44 | 93.6 | |
| Whisper CurriculumProtocol=Fine-tuned, Architecture=Whisper, Strategy=Curriculum Scheduling2026.06 | 30.52 | 7.85 | 93.4 | |
| W2V-BERT Curr.Protocol=Fine-tuned, Architecture=W2V-BERT, Strategy=Curriculum Scheduling2026.06 | 30.66 | 7.46 | 92.8 | |
| W2V-BERT MultilingualProtocol=Fine-tuned, Architecture=W2V-BERT, Strategy=Multilingual Baseline2026.06 | 30.89 | 6.83 | 93.2 | |
| Whisper Tone+Curr.Protocol=Fine-tuned, Architecture=Whisper, Strategy=Tone-conditioned + Curriculum2026.06 | 30.99 | 8.91 | 93.3 | |
| Whisper Tone-cond.Protocol=Fine-tuned, Architecture=Whisper, Strategy=Tone-conditioned2026.06 | 31.36 | 8.87 | 93.4 | |
| MMS CurriculumProtocol=Fine-tuned, Architecture=MMS, Strategy=Curriculum Scheduling2026.06 | 40.86 | 9.72 | 90.5 | |
| MMS MultilingualProtocol=Fine-tuned, Architecture=MMS, Strategy=Multilingual Baseline2026.06 | 43.34 | 10.25 | 89.9 | |
| MMS Tone+Curr.Protocol=Fine-tuned, Architecture=MMS, Strategy=Tone-conditioned + Curriculum2026.06 | 50.47 | 12.42 | 87.9 | |
| MMS Tone-cond.Protocol=Fine-tuned, Architecture=MMS, Strategy=Tone-conditioned2026.06 | 50.52 | 12.3 | 88 | |
| MMS-1B-AllProtocol=Zero-shot, Architecture=MMS2026.06 | 112.98 | 85.01 | 64.2 | |
| Whisper Large-v3-TurboProtocol=Zero-shot, Architecture=Whisper2026.06 | 146.3 | 71.66 | 67.6 |