XierHacker / Model_Fusion_Based_Prosody_Prediction

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-----------------------

1.run python convert.py

convert .utf-8 raw files to prosody tagged files

2.run python data_processing.py

trans prosody tagged files to dataset

-------------------use models to prediction-----------------

cd models

into models

run python bilstm_cbow.py

use bilstm_cbow to do prosody prediction

run python alignment.py

use alignment to do prosody prediction