Implementing Intonation Markup for a Concept-to-Speech System

Abstract:

In an earlier study Elke Teich and I compared two intonation markup systems. We compared the tone-based ToBI approach to the tune-based approach proposed by Systemic Functional Grammar. The goal of this comparison was to define a mapping between the two systems for the purpose of concept-to-speech generation in English. Although we made some interesting findings, one of the problems with this approach is that ToBI is essentially a phonological markup system rather than a phonetic one. It therefore doesn't provide an accurate description on the pitch contour. We are now investigating the Tilt model, and are planning to use this instead of ToBI.

In this talk I will give the background to the original study and argue why we hope the Tilt model will provide more useful results.

Bio:

Dr. Catherine Watson has a research background in speech processing, and acoustic phonetics. Her main research interests are acoustic analysis of New Zealand English, acoustic- articulatory modeling of speech, and intonation modeling. She was based in the Speech Hearing Language Research Centre from 1996-2001 but is now based in the Department of Electronics.

Enquiries: sals@mri.mq.edu.au