Streaming-attention
-
Provide codes of streaming attention mechanisms for end-to-end automatic speech recognition using Pytorch
-
Provide algorithms of the implementation of these streaming attention mechanisms for better understading
-
Compatible to ESPnet
Stable Monotonic Chunkwise Attention (sMoChA)
sMoChA is described in "Online Hybrid CTC/Attention Architecture for End-to-End Speech Recognition" by H Miao, G Gao, P Zhang, T Li, Y Yan, in Proceedings of Interspeech 2019.
Monotonic Truncated Attention (MTA)
MTA is described in "Online Hybrid CTC/attention End-to-End Automatic Speech Recognition Architecture" by H Miao, G Gao, P Zhang, Y Yan, in IEEE/ACM Transactions on Audio, Speech and Language Processing. Also see manuscript for brief description.