Printer Friendly

How to wreck a nice beach.

The way people pronounce words while reading aloud is often quite different from the way they say them in spontaneous speech. These differences may cause problems for companies trying to develop computers that recognize normal, continuous speech.

Usually, such machines are "trained" on speech samples read by a diverse group of people. However, if the machines were expected to understand spontaneous speech, many mismatches could occur. Current machines have difficulty distinguishing between rapidly spoken phrases like "how to recognize speech" and "how to wreck a nice beach."

At SRI International in Menlo Park, Calif., Jared Bernstein and Gay Baldwin are systematically studying the differences between spontaneous and prepared speech. In casual conversations, "you get all kinds of contractions of a type that aren't normally written down," says Bernstein. Often people don't even notice the contractions -- relaxed forms like "Idunno" and "gonta"--because they're listening for the content, he says.

So far, the researchers are finding that different people pronounce words differently and that the pronunciation often depends on the circumstances. They're also discovering that a person sometimes doesn't even have a consistent pronunciation for a given word. "Probably," for example, can sound like "pry," "probly," "probaly" or "prowubly."

"I was hoping that people would be consistently different," says Bernstein. Instead, people have surprisingly quirky and variable ways of pronouncing words. This means that machines that recognize and understand spontaneous speech may be even farther away than some researchers had hoped.
COPYRIGHT 1985 Science Service, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1985, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:problems in developing computers that recognize spontaneous speech
Author:Peterson, Ivars
Publication:Science News
Date:Nov 16, 1985
Previous Article:Brushing up on computer talk.
Next Article:The quality of violin strings.

Related Articles
Electronics recycling bill advances.
Speaking across the chasm: literature as a bridge between science and religion.
"Intelligent design," Natural Design, and the problem of meaning in the natural world.
Putting community in forests: a look back at the evolution of American Forests' policy niche and toward recommendations for expanding the role of...
Replanting after Katrina.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters