Multitask Learning with Local Attention for Tibetan Speech Recognition

Figure 6

The architecture of WaveNet-attention-CTC.