Paper: Improving Prosodic Boundaries Prediction for Mandarin Speech Synthesis by Using Enhanced Embedding Feature and Model Fusion Approach
Requirements
python3.5+
tensorflow>=1.6
numpy
pandas
scikit-learn
gensim
steps
----------------------data processing-----------------------
python convert.py
1.run convert
.utf-8
raw files to prosody tagged files
python data_processing.py
2.run trans prosody tagged files to dataset
-------------------use models to prediction-----------------
cd models
into models
python bilstm_cbow.py
run use bilstm_cbow to do prosody prediction
python alignment.py
run use alignment to do prosody prediction