The ESCA Workshop on Speech Synthesis

September 25-28, 1990
Autrans, France

Multi-Style Prosodic Model for French Text-To-Speech Synthesis

Valerie Pasdeloup

Universite du Québec, Institut National de la Recherche Scientifique, Ile-des-Soeurs, Québec, Canada

We present an attempt for model prosody, including a prosodic hierarchical organization and a series of linguistic and phonotatic rules. Unlike most other existing French speech synthesis systems where prosodic organization depends mainly on syntactic structure, we adopt an approach where the phonotatic constraints (on length of stress group, number of sequential unstressed syllables...) resulting biological and psychological prerequisites for speech activity are taken into account. In text-to-speech synthesis, our model proves effective in generating various acceptable prosodic structures for a given sentence.


Full Paper

Bibliographic reference.  Pasdeloup, Valerie (1990): "Multi-style prosodic model for French text-to-speech synthesis", In SSW1-1990, 193-196.