Attention Weight Smoothing Using Prior Distributions for Transformer-Based End-to-End ASR.

Takashi Maekaku, Yuya Fujita, Yifan Peng, Shinji Watanabe

Browse the full Interspeech paper archive.