ISCA Archive SpeechProsody 2004
ISCA Archive SpeechProsody 2004

Speech rate and prosody units: evidence of interaction from Mandarin Chinese

Chiu-yu Tseng, Yeh-lin Lee

This paper discusses evidence of interaction found between speech rate and prosody units in Mandarin Chinese speech. Mandarin speech data of 2 different speech rates that had been previously labeled for perceived boundaries and prosody units were further analyzed for duration patterns at each prosodic level. Each prosody level demonstrated patterns of duration adjustment for both speech rates that could be accounted for by the model used. These patterns of duration adjustments are clearly systematic, suggesting how each prosody levels may interact and to an extent govern the temporal distribution of units within. Our findings demonstrate that though speech rate may appear to be a global phenomenon across speech flow on the surface, it in fact is very much an in integrated part of prosody organization constrained by each prosody level. To put simply, duration adjustment is being made systematically at each prosody level during speech production instead of just an across-the-board phenomenon. As a result, interactions between prosody units and temporal distribution are predictable. We believe these findings are a step forward in understanding temporal organization and distribution of speech flow as well as speech prosody in general, and should be directly applicable to predicting speech prosody of unlimited TTS in particular.