An Integration of Pre-Trained Speech and Language Models for End-to-End Speech Recognition.
Yukiya HonoKoh MitsudaTianyu ZhaoKentaro MitsuiToshiaki WakatsukiKei SawadaPublished in: CoRR (2023)
Keyphrases
- speech recognition
- end to end
- language model
- pre trained
- language modeling
- speech signal
- n gram
- speech synthesis
- speech recognizer
- automatic speech recognition
- word error rate
- probabilistic model
- information retrieval
- retrieval model
- test collection
- speech recognition systems
- handwriting recognition
- query expansion
- speaker identification
- training data
- speech recognition technology
- mixture model
- training examples
- query terms
- noisy environments
- speaker dependent
- isolated word
- generative model
- semi supervised
- hidden markov models
- learning algorithm