Abstract
Text-to-speech systems are currently used in a variety of telephone applications for remote access to information. While this form of synthetic speech may be cost-effective relative to digitized speech, the impoverished quality of the speech signal may adversely affect its comprehensibility in telephone applications. The primary objective of the present research was to investigate the amount of familiarization needed to achieve an asymptotic level of comprehension performance with high-quality synthetic speech presented in the telephone environment. Sixty-four male and female native English speakers listened to digitized natural and digitized synthetic sentences that contained relatively high and low-predictable components. Subjects provided truth-value judgments for which accuracy, response time and response certainty were measured. Results indicate that a high-predictable introductory message of approximately three relatively short sentences may improve comprehension performance of high-quality synthetic speech in some telephone applications.
Get full access to this article
View all access options for this article.
