Automatic Speech Recognition on Cantonese JCCOCC MoCA (eval)
24.15CEREnc&Dec Prompts
Evaluation results
| Method | Links | |
|---|---|---|
| Enc&Dec PromptsModel=Whisper (LoRA)2026.06 | 24.15 | |
| Enc Only PromptsModel=Whisper (LoRA)2026.06 | 24.32 | |
| Audio-Textual PromptsModel=Whisper (LoRA)2026.06 | 24.78 | |
| Audio-Only PromptsModel=Whisper (LoRA)2026.06 | 25.32 | |
| Whisper (LoRA)Model=Whisper (LoRA)2026.06 | 25.76 | |
| Supervised (100%)Finetune Strategy=100% Sup.2026.06 | 25.79 | |
| RABModel=Whisper (LoRA)2026.06 | 26.55 | |
| Conformer Transducer with Only Speech ContextsModel=Conformer Transducer2026.06 | 26.96 | |
| x-vectorModel=Whisper (LoRA)2026.06 | 27.43 | |
| Incremental SAT systemFinetune Strategy=Semi-Sup., SAT=✓, TTA=✓2026.06 | 28.98 | |
| Incremental SI systemFinetune Strategy=Semi-Sup.2026.06 | 29.84 | |
| Semi-Supervised + SAT + TTAFinetune Strategy=Semi-Sup., SAT=✓, TTA=✓2026.06 | 30.17 | |
| ECAPA-TDNNModel=Whisper (LoRA)2026.06 | 30.19 | |
| Semi-Supervised (Random, Incr.)Finetune Strategy=Semi-Sup.2026.06 | 30.46 | |
| Semi-Supervised (Conf. Score)Finetune Strategy=Semi-Sup.2026.06 | 30.5 | |
| Semi-Supervised Baseline (Sys.3)Finetune Strategy=Semi-Sup.2026.06 | 30.86 | |
| Supervised (10%) + SAT + TTAFinetune Strategy=10% Sup., SAT=✓, TTA=✓2026.06 | 31.36 | |
| Supervised (10%)Finetune Strategy=10% Sup.2026.06 | 31.84 | |
| i-vectorModel=Whisper (LoRA)2026.06 | 36.16 |