Speech Recognition on German video dataset (clean)
11.2WER+ ce pretrain + ce loss, mid, top
Evaluation results
| Method | Links | ||
|---|---|---|---|
| + ce pretrain + ce loss, mid, toplambda_ce=0.62020.11 | 11.2 | — | |
| + aux losslambda_aux=0.32020.11 | 11.3 | — | |
| + aux + kl losslambda_aux=0.32020.11 | 11.3 | — | |
| + crosslingual pretrain + aux + kl losslambda_aux=0.32020.11 | 11.3 | — | |
| + ce pretrain + ce loss, midlambda_ce=0.62020.11 | 11.3 | — | |
| + crosslingual pretrain2020.11 | 11.4 | — | |
| + kl losslambda_aux=0.32020.11 | 11.5 | — | |
| + ce pretrain2020.11 | 11.5 | — | |
| baseline2020.11 | 11.6 | — |