Third Workshop on Spoken Language Technologies for Under-resourced Languages

Cape Town, South Africa
May 7-9, 2012

Modeling the Prosody of Vietnamese Attitudes for Expressive Speech Synthesis

Dang-Khoa Mac (1,2), Eric Castelli (1), Véronique Aubergé (2)

(1) International Research Institute MICA, HUST-CNRS/UMI 2954-Grenoble INP, Hanoi, Vietnam
(2) Laboratory of Informatics of Grenoble (LIG), CNRS, France

Attitudes or social affects are strongly implied in interaction processing, and specifically to socio-cultural aspects of language. This paper presents the modeling of attitude to apply in expressive speech synthesis in Vietnamese, an under-resourced tonal language. A prosodic model for Vietnamese attitude is proposed based on the concept of “rendez-vous” between linguistic levels and prosodic functions of utterance. This model is applied to generate the prosody of attitudes in Vietnamese. The perceptual experiment on the synthetic utterances with this model shows that the attitudes are well evaluated.

Index Terms: attitude, tone, prosodic modeling, expressive speech synthesis, Vietnamese

Full Paper

Bibliographic reference.  Mac, Dang-Khoa / Castelli, Eric / Aubergé, Véronique (2012): "Modeling the prosody of Vietnamese attitudes for expressive speech synthesis", In SLTU-2012, 114-118.