SGEM: Test-Time Adaptation for Automatic Speech Recognition via Sequential-Level Generalized Entropy Minimization.
Changhun KimJoonhyung ParkHajin ShimEunho YangPublished in: INTERSPEECH (2023)
Keyphrases
- automatic speech recognition
- entropy minimization
- speech recognition
- speech signal
- speaker adaptation
- hidden markov models
- broadcast news
- speech retrieval
- word error rate
- conversational speech
- recognition errors
- multiscale
- neural network
- word recognition
- spoken words
- haplotype inference
- spontaneous speech
- multi modal