Figure 1.
Figure 2.
Figure 3.
Figure 4.
Figure 5.
Figure 6.
Figure 7.
Train and test loss and ler of strategies based on accent
| Model | Test LER |
|---|---|
| Standard | 14.62% |
| Homogeneous gender | 6.71% |
| Heterogeneous gender | 13.13% |
| Homogeneous accent | 13.58% |
| Heterogeneous accent | 5.51% |
| Attention-LSTM[21] | 9.77% |
| CNN-LSTM[22] | 14.05% |
| BLSTM[23] | 12.9% |
| BiLSTM-E[24] | 8.07% |
Loss, LER for accent-based strategies during both train and test
| Test Loss | Test LER | Train Loss | Train LER | |
|---|---|---|---|---|
| standard | 25.69 | 14.62% | 12.72 | 11.14% |
| Homogeneous accent | 22.26 | 13.58% | 11.64 | 10.04% |
| Heterogeneous accent | 12.08 | 5.51% | 5.70 | 4.42% |
Loss, LER for gender-based strategies during both train and test
| Test Loss | Test LER | Train Loss | Train LER | |
|---|---|---|---|---|
| Standard | 25.69 | 14.62% | 12.72 | 11.14% |
| Homogeneous gender | 13.53 | 6.71% | 6.15 | 5.65% |
| Heterogeneous gender | 24.15 | 13.12% | 11.08 | 9.21% |