Prediction of Prosodic Word Boundaries in Chinese TTS Based on Maximum Entropy Markov Model and Transformation Based Learning

Hierarchical prosody structure generation is a key component for a speech synthesis system. As the basic prosodic unit, the prosodic word plays an important role for the naturalness and the intelligibility for the Chinese TTS system. In this paper we proposed an approach for prediction of Chinese prosodic word boundaries in unrestricted Chinese text, which combines Maximum Entropy Markov Model(MEMM) and TBL model. First MEMM is trained to predict the prosodic word boundaries. After that we apply a TBL based error driven learning approach to amend the initial prediction. A comparison is conducted between the new model and HMM for prosodic word boundaries prediction. Experiments show that the combined approach improves overall performance. The precision and recall ratio are improved.