Endpoint Detection on Internal Speech Corpus
5EINext-Turn (Whisper-large)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Next-Turn (Whisper-large)Param. (M)=640, Lat. (ms)=1522026.06 | 5 | 86.7 | 88.4 | |
| Next-Turn (Whisper-small)Param. (M)=89, Lat. (ms)=522026.06 | 10.7 | 81.1 | 83.5 | |
| Next-Turn (Whisper-tiny)Param. (M)=8, Lat. (ms)=232026.06 | 12.3 | 73.2 | 78.2 | |
| Acoustic VAD (Silero)threshold=640 ms, Param. (M)=0.5, Lat. (ms)=32026.06 | 27 | 5 | 61.6 | |
| Easy TurnParam. (M)=850, Lat. (ms)=2632026.06 | 31.1 | 60.8 | 67.8 | |
| Acoustic VAD (Silero)threshold=480 ms, Param. (M)=0.5, Lat. (ms)=32026.06 | 45.7 | 10.4 | 52.5 | |
| Smart Turn v2Param. (M)=95, Lat. (ms)=292026.06 | 59.6 | 32.5 | 35.2 | |
| Acoustic VAD (Silero)threshold=320 ms, Param. (M)=0.5, Lat. (ms)=32026.06 | 60 | 33.9 | 39.7 | |
| Smart Turn v3.2Param. (M)=8, Lat. (ms)=212026.06 | 64.5 | 30.5 | 35.4 |