STiLL - Speech Technology in Language Learning

May 25-27, 1998
Marholmen, Sweden

Models of Australian English Vowels for Commercial Visual Feedback Systems

J. van Doorn, J. Shakeshaft, A. Winkworth, L. Hand, S. Joshi

School of Communication Sciences and Disorders, The University of Sydney, Sydney, Australia

Visual feedback via computer-based speech technology systems is gaining momentum in assisting with correct pronunciation in second language learning [1]. Two commercial systems, the IBM SpeechViewer and the Sonamatch module of the Kay CSL system use spectral matching of target speech models to provide feedback to clients about their pronunciations. This study explored the option of using stored general models of Australian English vowels as representative targets in both SpeechViewer and Sonamatch. Two groups of adult native speakers of Australian English (8 men and 10 women) were used to generate a set of integrated spectral models for each group. These models underwent preliminary testing for their ability to recognise vowels produced by a native speaker of Australian English and a speaker of American English. Implications for the use of general dialectal models of vowels in visual feedback systems as an aid to correct pronunciation of vowels are discussed.

Reference

  1. ึster A-M (1997) Auditory and Visual Feedback in Spoken L2 Teaching, Phonum, 4: 145-148.

Full Paper

Bibliographic reference.  Doorn, J. van / Shakeshaft, J. / Winkworth, A. / Hand, L. / Joshi, S. (1998): "Models of Australian English vowels for commercial visual feedback systems", In STiLL-1998, 53-56.