1.RNN,Seq2Seq,Attention简介：
完全图解RNN、RNN变体、Seq2Seq、Attention机制
补充知识：LSTM

2.Attention存在的问题：
(I)对噪声敏感
(II)对长句识别性能下降
GMIS 2017 | 腾讯AI Lab副主任俞栋：语音识别研究的四大前沿方向

3.解读transformer：
Attention is All You Need | 每周一起读

4.解读CNN机器翻译：
Facebook提出全新CNN机器翻译：准确度超越谷歌而且还快九倍

1.attention用于语音识别前的baseline（用于翻译），对于长句识别效果急剧变差：
Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. Neural machine translation by jointly learning to align and translate. arXiv:1409.0473, September 2014

2.第一篇将attention用于语音识别的文章(提出hybird attention机制，解决attention的位置信息问题，提出attention加窗机制，数据集：TIMIT)：
J.Chorowski, D.Bahdanau, D.Serdyuk, K.Cho, and Y.Bengio."Attention-based models for speech recognition"

3.attention加窗改进，RNN更换为GRU在LVCSR任务上的应用（数据集：(WSJ) corpus (available as LDC93S6B and LDC94S13B)）：
D. Bahdanau, J. Chorowski, D. Serdyuk, P. Brakel, and Y. Bengio,“End-to-end attention- based large vocabulary speech recognition,”

4.最经典的attention语音识别模型：
W. Chan, N. Jaitly, Q. Le, and O. Vinyals, “Listen, attend and spell: Aneural network for large vocabulary conversational speech recognition”

5.CTC+Attention(CTC强化了对齐的单调性，且CTC加快网络训练速度，数据集：WSJ1 (81 hours) ,WSJ0 (15 hours) , CHiME-4 (18 hours)):
S. Kim, T. Hori, and S. Watanabe, “Joint CTC-attention based end-to-end speech recognition using multi-task learning”

6.transformer:
N. Shazeer, Niki Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, I. Polosukhin "Attention Is All You Need"

7.CNN机器翻译：
J. Gehring, M. Auli, D. Grangier, D. Carats, Y. N. Dauphin "Convolutional Sequence to Sequence Learning"

8.CNN+RNN的E2E ASR：
VERY DEEP CONVOLUTIONAL NETWORKS FOR END-TO-END SPEECH RECOGNITION

9.CNN+RNN attention+CTC（Encoder在RNN前加入VGGnet进行特征提取，数据集：WSJ0,CHIME-4）:
Kim S, Hori T, Watanabe S. Joint CTC-attention based end-to-end speech recognition using multi-task learning

10.self-attention+CTC:
SELF-ATTENTION NETWORKS FOR CONNECTIONIST TEMPORAL CLASSIFICATION IN SPEECH RECOGNITION

1.tensorflow E2EASR代码

2.pytorch E2EASR代码
需要解决的问题：如何多GPU运行