Domain-optimized Chinese speech generation.

Fung Tien Ying.Thesis (M.Phil.)--Chinese University of Hong Kong, 2001.Includes bibliographical references (leaves 119-128).Abstracts in English and Chinese.Abstract --- p.1Acknowledgement --- p.1List of Figures --- p.7List of Tables --- p.11Chapter 1 --- Introduction --- p.14Chapter 1.1 --- General Trends on Speech Generation --- p.15Chapter 1.2 --- Domain-Optimized Speech Generation in Chinese --- p.16Chapter 1.3 --- Thesis Organization --- p.17Chapter 2 --- Background --- p.19Chapter 2.1 --- Linguistic and Phonological Properties of Chinese --- p.19Chapter 2.1.1 --- Articulation --- p.20Chapter 2.1.2 --- Tones --- p.21Chapter 2.2 --- Previous Development in Speech Generation --- p.22Chapter 2.2.1 --- Articulatory Synthesis --- p.23Chapter 2.2.2 --- Formant Synthesis --- p.24Chapter 2.2.3 --- Concatenative Synthesis --- p.25Chapter 2.2.4 --- Existing Systems --- p.31Chapter 2.3 --- Our Speech Generation Approach --- p.35Chapter 3 --- Corpus-based Syllable Concatenation: A Feasibility Test --- p.37Chapter 3.1 --- Capturing Syllable Coarticulation with Distinctive Features --- p.39Chapter 3.2 --- Creating a Domain-Optimized Wavebank --- p.41Chapter 3.2.1 --- Generate-and-Filter --- p.44Chapter 3.2.2 --- Waveform Segmentation --- p.47Chapter 3.3 --- The Use of Multi-Syllable Units --- p.49Chapter 3.4 --- Unit Selection for Concatenative Speech Output --- p.50Chapter 3.5 --- A Listening Test --- p.51Chapter 3.6 --- Chapter Summary --- p.52Chapter 4 --- Scalability and Portability to the Stocks Domain --- p.55Chapter 4.1 --- Complexity of the ISIS Responses --- p.56Chapter 4.2 --- XML for input semantic and grammar representation --- p.60Chapter 4.3 --- Tree-Based Filtering Algorithm --- p.63Chapter 4.4 --- Energy Normalization --- p.67Chapter 4.5 --- Chapter Summary --- p.69Chapter 5 --- Investigation in Tonal Contexts --- p.71Chapter 5.1 --- The Nature of Tones --- p.74Chapter 5.1.1 --- Human Perception of Tones --- p.75Chapter 5.2 --- Relative Importance of Left and Right Tonal Context --- p.77Chapter 5.2.1 --- Tonal Contexts in the Date-Time Subgrammar --- p.77Chapter 5.2.2 --- Tonal Contexts in the Numeric Subgrammar --- p.82Chapter 5.2.3 --- Conclusion regarding the Relative Importance of Left versus Right Tonal Contexts --- p.86Chapter 5.3 --- Selection Scheme for Tonal Variants --- p.86Chapter 5.3.1 --- Listening Test for our Tone Backoff Scheme --- p.90Chapter 5.3.2 --- Error Analysis --- p.92Chapter 5.4 --- Chapter Summary --- p.94Chapter 6 --- Summary and Future Work --- p.95Chapter 6.1 --- Contributions --- p.97Chapter 6.2 --- Future Directions --- p.98Chapter A --- Listening Test Questionnaire for FOREX Response Genera- tion --- p.100Chapter B --- Major Response Types For ISIS --- p.102Chapter C --- Recording Corpus for Tone Investigation in Date-time Sub- grammar --- p.105Chapter D --- Statistical Test for Left Tonal Context --- p.109Chapter E --- Statistical Test for Right Tonal Context --- p.112Chapter F --- Listening Test Questionnaire for Backoff Unit Selection Scheme --- p.115Chapter G --- Statistical Test for the Backoff Unit Selection Scheme --- p.117Chapter H --- Statistical Test for the Backoff Unit Selection Scheme --- p.118Bibliography --- p.11