Third International Conference on Spoken Language Processing (ICSLP 94)

Yokohama, Japan
September 18-22, 1994

A Feature-Profile for Application-Specific Speech Synthesis Assessment and Evaluation

Ute Jekosch (1), Louis C. W. Pols (2)

(1) Lehrstuhl für allgemeine Elektrotechnik und Akustik, Ruhr-Universität Bochum, Bochum, Germany
(2) Institute of Phonetic Siences / IFOTT, University of Amsterdam, Amsterdam, The Netherlands

This paper is concerned with assessing the quality of speech synthesizers in application utilizing the conclusions made by subjects. Ultimately, the quality of speech becomes apparent in its usage, in this case the interaction between a machine and a biological system in a specific communicative context. As it is not possible to attain precise control over environmental circumstances of system application, however, a vast range of complicated human responses is possible. The approach introduced here is an attempt to bridge the gap between theory and experiment on the one hand, and, the actual performance of the system in application on the other hand, while still advocating the opinion that laboratory analogues of real applications are conceivable. What is still missing at present is a meaningful correlation between the properties of the speech signal and their relative relevance in an application-specific pragmatic context. Such a functionalized view of speech requires a unique formal description of both the speech signal properties and independently defined rudimentary variables characteristic of the situational communicative context.

Full Paper

Bibliographic reference.  Jekosch, Ute / Pols, Louis C. W. (1994): "A feature-profile for application-specific speech synthesis assessment and evaluation", In ICSLP-1994, 1319-1322.