Automatic Speech Recognition on Hokkien
21.34CERZipformer
Evaluation Results
| Method | Links | |
|---|---|---|
| ZipformerTraining Data=ASR+ST, Training Steps=320K, Decoding Strategy=Greedy2026.01 | 21.34 | |
| ZipformerTraining Data=ASR-only, Training Steps=312K, Decoding Strategy=Greedy2026.01 | 21.66 | |
| FireRed-AED2026.01 | 88.65 | |
| Paraformer2026.01 | 90.63 |