Speech-Transformer: A No-Recurrence Sequence-to-Sequence Model for Speech Recognition
Recurrent sequence-to-sequence models using encoder-decoder architecture have made great progress in speech recognition task. However, they suffer from the drawback of slow trai...