Second International Conference on Spoken Language Processing (ICSLP'92)

Banff, Alberta, Canada
October 13-16, 1992

Yet Another Rule Compiler for Text-To-Speech Conversion?

Ken Ceder, Bertil Lyberg

Telia Research AB, Systems Research, Spoken Language Processing, Haninge, Sweden

An interactive environment for developing applications and conducting research in the domain of text-to-speech conversion has been designed. The system is implemented in Prolog and is a new generation of the subclass of speech synthesis-by-rule systems, called rule compilers. It integrates different levels of linguistic and acoustic analyses and is constructed specifically for linguists and phoneticians. Special notational formalisms have been devised, which as closely as possible correspond to phonetic and linguistic terminology. Two types of synthesis strategies are included in the system, the formant synthesis and the polyphone synthesis, i.e., concatenation of speech units of arbitrary size. In the present paper it is shown how this system can be used for implementing a text-to-speech synthesis for Swedish.

Full Paper

Bibliographic reference.  Ceder, Ken / Lyberg, Bertil (1992): "Yet another rule compiler for text-to-speech conversion?", In ICSLP-1992, 1151-1154.