Improving Naturalness and Controllability of Sequence-to-Sequence Speech Synthesis by Learning Local Prosody Representations.
Cheng GongLongbiao WangZhenhua LingShaotong GuoJu ZhangJianwu DangPublished in: ICASSP (2021)
Keyphrases
- speech synthesis
- text to speech
- learning process
- learning algorithm
- reinforcement learning
- learning tasks
- external representations
- hidden state
- word processing
- action sequences
- inductive inference
- speech recognition
- learning systems
- unsupervised learning
- prior knowledge
- feature selection
- background knowledge
- learning scenarios
- online learning
- supervised learning
- active learning
- partial observability
- pattern recognition
- neural network