Sixth ISCA Workshop on Speech Synthesis

Bonn, Germany
August 22-24, 2007

Data-Driven Approach to Rapid Prototyping Xhosa Speech Synthesis

Justus C. Roux, Albert S. Visagie

Centre for Language and Speech Technology, Stellenbosch University, South Africa

This paper presents work in progress towards building a Xhosa speech synthesizer. HTS is being used for this purpose due to certain desirable properties. As a minority language, linguistic resources for Xhosa are limited despite a variety of impressionistic phonetic studies, prompting a minimalist approach and a preference for data-driven methods. Xhosa is an agglutinative language, and is also held to be a tonal language, which therefore requires morphological analysis and tonal information in order to generate intelligible speech. By taking into account more recent findings on the nature of Xhosa prosody, it appears that a minimalist approach that excludes tone information is possible. We implement the system using HTS. Such a data-driven TTS system is a useful tool to test various syntactic and other features in text that influence Xhosa prosody.

Full Paper      Presentation (ppt)    Sound examples  

Bibliographic reference.  Roux, Justus C. / Visagie, Albert S. (2007): "Data-driven approach to rapid prototyping Xhosa speech synthesis", In SSW6-2007, 143-147.