This is an ROW/Planning Grant. In recent years, work in speech act theory has yielded results of practical interest, particularly for the design of user interfaces, intelligent systems, and programming languages. Techniques have been developed for general reasoning about actions, and for inference of intentions from typewritten dialogue. This project adds intonation to the model's language cues. A parametric study of spoken language use will be conducted, to support the hypothesisthat computing intentions is improved by exploiting acoustic parameters (particularly the fundamental frequency contours) of utterances. Speakers' reports of their own intentions, and hearers' judgements of intent will be used, to determine the roleof intonation relative to linguistic structure and contextual cues. The work will attempt to develop a calculus relating international features and speech act types, and fold it into the existing speech act recognition system. Acoustic cues may provide new insight into the dynamic nature of action and communication.
