Speech representation learning-paper

• Disentanglement papers
  • Streaming ASR
    • RNA based
    • RNN-T based
    • Attention based
    • Unified Streaming/Non-streaming models
  • Non-autoregressive (NAR) ASR
  • ASR Rescoring / Spelling Correction (2-pass decoding)
  • On-device ASR

Disentanglement Papers

List of papers in disentanglment or disentangled representation learning. I am working the disentanglement in VC(voice conversion). I have met a lot of related papers and find it annoying to orginize on my MAC. That's why this repo comes. I did not do much research on disentanglement and know a little about it, so there might be some mistakes. Welcome to correct!

If you are interested in this repo, any pull request is welcomed.

VAE

limit bottleneck capacity

a probabilistic version( slow feature analysis)

• Slow VAE:Towards Nonlinear Disentanglement in Natural Data with Temporal Sparse Coding(ICLR 2021)
• \beta VAE:PREVENTING POSTERIOR COLLAPSE WITH δ-VAES(ICLR 2019)

Speech Representation Learning Papers

Universal representation

• SpeechNet
• SpeechT5

CPC

• wav2vec-c: Wav2vec-C: A Self-supervised Model for Speech Representation Learning

APC

Generative pre-training for speech with autoregressive predictive coding（昨天和xi wang老师还聊到表征学习、预训练这个问题）

For VC

• Pretraining for VC (https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9314100)