Third International Conference on Spoken Language Processing (ICSLP 94)

Yokohama, Japan
September 18-22, 1994

Towards a Prosodic Cues-Based Modelling of Phonological Variability for Text-to-Speech Synthesis

Anne Lacheret-Dujour (1,2), Vincent Pean (1)

(1) LIMSI-CNRS BP133, Orsay-Cedex, France
(2) ELSAP, Caen Cedex, France

The study presented in this paper has been carried out in the framework of phonological variability in French, with applications in automatic speech processing in mind. The specific aim of our study is to model inter-speaker phonemic variants at word boundaries: global strategies and strategies specific to different classes of speakers. Data have been collected from casual speech corpus. Variants are defined through their relationship with the prosodic structure of a given utterance, This structure depends on syntactic organization of discourse. One example of this phenomena is given here with the pronunciation of the schwa at words' frontiers.

Full Paper

Bibliographic reference.  Lacheret-Dujour, Anne / Pean, Vincent (1994): "Towards a prosodic cues-based modelling of phonological variability for text-to-speech synthesis", In ICSLP-1994, 1763-1766.