Automatic Speech Recognition on NER-MIT-OpenCourseWare (test)
22.7NE WERWhisper-la (Proposed Method)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Whisper-la (Proposed Method)Params=1.54 B, Revision Status=After revision, Revision LLM=GPT-4o-mini2025.06 | 22.7 | 7 | |
| Whisper-me (Proposed Method)Params=769 M, Revision Status=After revision, Revision LLM=GPT-4o-mini2025.06 | 26.8 | 8 | |
| Canary-1B (Proposed Method)Params=1B, Revision Status=After revision, Revision LLM=GPT-4o-mini2025.06 | 26.8 | 8.3 | |
| Whisper-sm (Proposed Method)Params=244 M, Revision Status=After revision, Revision LLM=GPT-4o-mini2025.06 | 29.5 | 9.4 | |
| Whisper-laParams=1.54 B, Revision Status=Before revision2025.06 | 32.3 | 7 | |
| Whisper-meParams=769 M, Revision Status=Before revision2025.06 | 36 | 8 | |
| Canary-1BParams=1B, Revision Status=Before revision2025.06 | 36.6 | 8.3 | |
| Whisper-smParams=244 M, Revision Status=Before revision2025.06 | 39.3 | 9.4 |