Sixth European Conference on Speech Communication and Technology

Budapest, Hungary
September 5-9, 1999

Canadian French Text-to-Speech Synthesis: Modeling an Optimal Set of Realizations for Dialect Markers

Evelyne Tzoukermann, Lucie Ménard, Marise Ouellet

Bell Laboratories Lucent Technologies, Murray Hill, NJ, USA

We report on the development of a text-to-speech system for Canadian French which builds on both our theoretical and practical experience from the European French system [8, 10]. In order to extend to an additional dialect, we took an approach which would permit us to rapidly build the new TTS system. Specifically, we (1) identified speech markers for the new dialect, (2) determined optimal implementation points for these markers, at the phone inventory, diphone recording, text analysis, and duration levels. Potential units were filtered based on generalizable spectral properties and on frequency of occurrence.

Full Paper (PDF)   Gnu-Zipped Postscript

Bibliographic reference.  Tzoukermann, Evelyne / Ménard, Lucie / Ouellet, Marise (1999): "Canadian French text-to-speech synthesis: modeling an optimal set of realizations for dialect markers", In EUROSPEECH'99, 2119-2122.