Fig. 4From: Towards end-to-end speech recognition with transfer learningJoint CTC-attention model with a shallow encoder on high-level featuresBack to article page