Rhythm and stress-shift in speech synthesis

Abstract

This paper presents the basic phonological rhythm rule used in the intonation component of the Edinburgh University Centre for Speech Technology Research (CSTR) text-to-speech system, and shows how stress-shift phenomena can be incorporated into such a rule. The first part of the paper discusses the need for rhythm rules in speech synthesis and outlines the operation of the CSTR rule. The second part discusses the implications of such a rule for lexical stress and the applicability of its output to synthesis by rule.