ASR reliability selection on JASMIN Dutch Dialogue Material
97.4PrecisionAgreement Whisper-V2 and Whisper-FT
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Agreement Whisper-V2 and Whisper-FTSelection Logic=[LLM-classification], LLM Model=ChatGPT-5 (gpt-5-2025-08-07)2026.04 | 97.4 | 53.8 | 69.3 | 45 | |
| Whisper-FTSelection Logic=[LLM-classification], LLM Model=ChatGPT-5 (gpt-5-2025-08-07)2026.04 | 88.9 | 93.8 | 91.3 | 62 | |
| Whisper-V2Selection Logic=[LLM-classification], LLM Model=ChatGPT-5 (gpt-5-2025-08-07)2026.04 | 62.1 | 95.6 | 75.3 | 48 |