A Maximum Entropy Markov Model for Prediction of Prosodic Phrase Boundaries in Chinese TTS

Abstract

Hierarchical prosody structure generation is a key component for a speech synthesis system. One major feature of the prosody of Mandarin Chinese speech flow is prosodic phrase grouping. In this paper a method based on maximum entropy Markov model (MEMM) is proposed to predict prosodic phrase boundaries in unrestricted Chinese text. MEMM is described in detail that combines transition probabilities and conditional probabilities of states effectively. The conditional probabilities of states are estimated by maximum entropy (ME) theory. A comparison is conducted between the new model and maximum entropy model for prosody phrase break prediction. The experiments show that utilizing the same feature set, MEMM improves overall performance. The precision and recall ratio are improved.