Semi-Supervised Learning Based on Hierarchical Generative Models for End-to-End Speech Synthesis.
Takato FujimotoShinji TakakiKei HashimotoKeiichiro OuraYoshihiko NankakuKeiichi TokudaPublished in: ICASSP (2020)
Keyphrases
- end to end
- semi supervised learning
- speech synthesis
- discriminative models
- generative model
- semi supervised
- speech recognition
- unsupervised learning
- unlabeled data
- labeled data
- text to speech
- mixture model
- machine learning
- probabilistic model
- supervised learning
- discriminative learning
- active learning
- training data
- co training
- transfer learning
- pairwise
- text categorization
- topic models
- multi view
- conditional random fields
- real world
- class labels
- prior knowledge
- pairwise constraints
- text classification
- hidden markov models
- sparse coding
- information extraction