Towards Spontaneous Style Modeling with Semi-supervised Pre-training for Conversational Text-to-Speech Synthesis.
Weiqin LiShun LeiQiaochu HuangYixuan ZhouZhiyong WuShiyin KangHelen MengPublished in: CoRR (2023)
Keyphrases
- semi supervised
- text to speech synthesis
- supervised learning
- active learning
- supervised methods
- training process
- feature selection
- conversational speech
- semi supervised classification
- training phase
- co training
- labeled data
- online learning
- neural network
- multiple instance learning
- modeling method
- weakly supervised
- multi modal
- restricted boltzmann machine
- training set
- fully labeled