Fourth European Conference on Speech Communication and Technology

Madrid, Spain
September 18-21, 1995

The generation of prosody in the Nijmegen Rule Oriented Speech synthesis system

J. Kerkhoff, T. Rietveld

University of Nijmegen, Dept. of Language and Speech, Nijmegen, The Netherlands

In this contribution a description is given of the modules developed for implementing Dutch prosody, and more in particular an autosegmental description of Dutch intonation in the rule-based Text-to-Speech system NIROS: Nijmegen Interactive Rule Oriented Speech synthesis. Our TTS systeem is completely syntax directed; the synthesizer part is based on a serial pole-zero (ARMA) structure [9]. In NIROS a multi-layer datastructure [6,12] is used reflecting the intuitive notion that speech is organized at serveral levels both segmental and suprasegmental. By letting rules work on the different layers (datastre-ams) of this datastructure the input of the synthesizer is prepared. Emphasis is given both to the multi-layered datastructure, and to the two compilers used in NIROS to manipulate data: the FONPARS rule compiler for handling monolayered datastructures, and the compiler ALF-EIOS for multilayered data. Keywords: text-to-speech, prosody

Full Paper

Bibliographic reference.  Kerkhoff, J. / Rietveld, T. (1995): "The generation of prosody in the nijmegen rule oriented speech synthesis system", In EUROSPEECH-1995, 1831-1834.