LET'S GO: Improved Speech Interfaces For The General Public
Carnegie Mellon University, Pittsburgh PA
Investigators
Abstract
This project investigates spoken language systems for the elderly and non-native speakers of English. Existing systems have been constructed for the average user and thus perform poorly for these nontraditional populations. The elderly have little access at present to computerized information, yet they have telephones that would enable them to access computers easily if the information were presented in a way that they could readily understand. Non-native speakers have difficulty being understood by the same dialogue systems, which need to be adapted to their speech and aid them in finding the correct way to express themselves. In order to improve the understandability of spoken output, Let's Go will investigate differences in lexical, prosodic and spectral content to enable elderly and non-native listeners to better understand the message. In order to improve spoken input, the project will explore adaptation of the statistical models of the lexical and acoustic information of the incoming speech signal for the target populations. Let's Go will assess the pertinence of its findings by creating a spoken language demonstrator that allows elderly and non-native users to access Port Authority Transit Bus information over the telephone.
View original record on NSF Award Search →