A study on form and function of prosody based on acoustics, interpretation, and modelling - with evidence from the analysis by synthesis of Mandarin speech prosody

An analysis-by-synthesis study on Mandarin speech prosody is conducted in the present dissertation. The features of Mandarin speech prosody are discussed by focusing on two salient aspects: the function of prosody and the form of prosody. The study attempts to find a plausible way in which the two aspects can be mapped onto each other through the functional analysis of prosody and the multi-level formal representation. The form of Mandarin speech prosody is a complex F0 picture due to the simultaneous uses of pitch contours by both lexical tones and sentential intonation. The phenomenon of tone sandhi in speech context triggers more puzzling issues when researchers are confronted with the acoustic form of Mandarin prosody. The functional use of prosody in Mandarin speech concerns: at the lexical level for word identity (Tone1, Tone2, Tone3, Tone4, and Tone0); at the sentential level for prominence marking (sentence accents) and the indication of prosodic boundaries (intonation boundary tones). In the present study, the analysis of prosodic function at the two levels provides a basic framework in coding the surface melodic form of Mandarin prosody, which consists of pitch contours in tonal units and boundary tones at the beginning and end of intonation unit. For the formal representation of Mandarin speech prosody, the surface F0 contour of each utterance is coded into a sequence of INTSINT symbols, and subject to the Prozed tool for speech synthesis. It is shown that the synthesized stimuli derived from the symbolic coding can closely follow the melodic features and correctly express the prosodic function of the original Mandarin utterances. The present study employs acoustic data, symbolic coding, and speech synthesis for the derivative mapping between prosodic function and form, which aims to interpret the complex prosodic phenomenon, and provide an insight for the annotation and analysis of Mandarin speech prosody.