Self-Training and Pre-Training are Complementary for Speech Recognition.
Qiantong XuAlexei BaevskiTatiana LikhomanenkoPaden TomaselloAlexis ConneauRonan CollobertGabriel SynnaeveMichael AuliPublished in: ICASSP (2021)
Keyphrases
- speech recognition
- wall street journal corpus
- isolated word
- training set
- hidden markov models
- acoustic models
- speech understanding
- speech recognizer
- speech signal
- speech synthesis
- automatic speech recognition
- language model
- pattern recognition
- speech processing
- speaker identification
- speech recognition technology
- noisy environments
- speech recognition errors
- co training
- training process
- keyword spotting
- speech recognition systems
- speech retrieval
- semi supervised learning
- supervised learning
- speech recognizers
- active learning