Continual Self-Supervised Domain Adaptation for End-to-End Speaker Diarization.
Juan Manuel CoriaHervé BredinSahar GhannaySophie RossetPublished in: SLT (2022)
Keyphrases
- end to end
- domain adaptation
- speaker diarization
- cross domain
- speech recognition
- multiple sources
- transfer learning
- semi supervised
- labeled data
- semi supervised learning
- target domain
- sentiment classification
- broadcast news
- bayesian information criterion
- unlabeled data
- test data
- machine learning
- document classification
- data sets
- data sources
- active learning
- training data
- computer vision