Table 8 Comparisons on different heads for multi-level attention
From: A new joint CTC-attention-based speech recognition model with multi-level multi-head attention
| Â | 2 heads | 3 heads | 4 heads | 5 heads |
|---|---|---|---|---|
TIMIT | 16.73 | 16.52 | 16.34 | 16.60 |
WSJ | 4.2 | 4.0 | 3.8 | 4.0 |
LibriSpeech | 3.9 | 3.7 | 3.6 | 3.8 |