Improving non-autoregressive end-to-end speech recognition with pre-trained acoustic and language models.
Keqi DengZehui YangShinji WatanabeYosuke HiguchiGaofeng ChengPengyuan ZhangPublished in: CoRR (2022)
Keyphrases
- end to end
- speech recognition
- language model
- autoregressive
- pre trained
- speech recognition systems
- speaker independent
- language modeling
- non stationary
- random fields
- n gram
- probabilistic model
- information retrieval
- speech signal
- training data
- sar images
- automatic speech recognition
- word error rate
- training examples
- hidden markov models
- optical flow
- bayesian networks
- pattern recognition
- co occurrence
- conditional random fields
- supervised learning
- image segmentation
- multimedia