Speech Recognition on Switchboard Eval2000
9SW Error RateTDNN-LSTM-D
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| TDNN-LSTM-D2020.05 | 9 | — | 13.9 | |
| TDNN-D2020.05 | 9.6 | 19.9 | 14.8 | |
| CTC-CRF (CSF)Model=chunk-based with context + twin reg2020.05 | 9.7 | 18.4 | 14.1 | |
| CTC-CRF (CSF)Model=chunk-based with context + chunk size jitter2020.05 | 10.4 | 19.5 | 15 | |
| CTC-CRF (CSF)Model=chunk-based with context2020.05 | 10.7 | 20 | 15.4 | |
| CTC-CRF (SF)Model=chunk-based w/o future context + twin reg2020.05 | 10.8 | 19.7 | 15.3 | |
| CTC-CRF (SF)Model=chunk-based w/o future context2020.05 | 11 | 20.4 | 15.7 | |
| CTC-CRF (SF)Model=chunk-based w/o future context + chunk size jitter2020.05 | 11 | 19.7 | 15.4 | |
| online-enabled BLSTM2020.05 | 11.6 | 23 | 17.3 |