Multitask Learning with Local Attention for Tibetan Speech Recognition

Figure 5

The architecture of attention-WaveNet-CTC.