Speech Synthesis by Rule as a Research Technique

Publisher Website
Google Scholar
Abstract
A number of investigators, using various types of systems, have for some time been engaged in generating synthetic speech from a phonemic input. This work is not just an interesting curiosity but has practical applications and is a tool for linguistic research. Reasonably intelligible speech can now be produced, not only because of our growing understanding of the problems involved, but also because of technical improvements in synthesizers and the increased flexibility in experimentation that digital computers have made possible. But comparison of speech synthesized by rule with closely copied speech produced on the same system makes clear that the rules still need a great deal of correction and, perhaps, further elaboration. Areas in which further work is particularly needed include the prosodic features, rhythm and duration, “naturalness,” allophonic rules, and juncture. All of these topics are of great interest to phoneticians and phonologists. Speech synthesis by rule provides both an intellectual framework for consideration of these problems and an experimental method with which to investigate them.

This publication has 0 references indexed in Scilit: