论文信息 - Automatic Detection of Prosody Phrase Boundaries for Text-to-Speech System

Automatic Detection of Prosody Phrase Boundaries for Text-to-Speech System

Automatic acquisition of the prosodic phrase boundary detecting rules from the text and speech corpora has always been a difficulty for TTS systems. We collected over 5,000 sentences as the corpus, introduced a method based on the transform-based error-driven learning to get the rules for detecting prosodic phrase boundaries, and then used trees to organize the rules in the TTS system. For using the transformation-based error-driven learning, we designed a set of templates especially. Using 1,000 sentences to get rules for the TTS system can reach 92% accuracy in close-test and 73% accuracy in open-test.

Tiejun Zhao | Muyun Yang | Xin Lv | Zhanyi Liu

[1] Kuldip K. Paliwal,et al. Speech Coding and Synthesis , 1995 .

[2] Yung-Hwan Oh,et al. Tree-based modeling of prosodic phrasing and segmental duration for Korean TTS systems , 1999, Speech Commun..

[3] Zhao Tie. Increasing Accuracy of Chinese Segmentation with Strategy of Multi step Processing , 2001 .

[4] Eric Brill,et al. A corpus-based approach to language learning , 1993 .

[5] Eric Brill,et al. Transformation-Based Error-Driven Learning and Natural Language Processing: A Case Study in Part-of-Speech Tagging , 1995, CL.

[6] Rpg Rene Collier,et al. The generation of prosodic structure and intonation in speech synthesis , 1995 .