REPETITION AND RE-START STRATEGIES FOR PROSODY IN TEXT-TO-SPEECH CONVERSION SYSTEMS

Speakers in conversations between humans continually adapt the prosodic and structural aspects of their speech to the perceived needs of their listeners, in terms of judgments about the potentially masking effects of transient and ambient noise levels, and in response to explicit requests by listeners for repetition. Adaptive strategies for repetition include changing such prosodic aspects of utterances as pitch range and mean, intensity mean and overall tempo of speaking, together with intonational re-structuring. Such repetition also deploys re-start strategies based on structural linguistic knowledge. An outline is offered of principles for incorporating elements of such intelligent adaptivity in the operation of text-to-speech conversion systems, to improve their interactive ability with human partners in dialogue.