Towards realistic verbal interactions between people and computers -- A probabilistic approach

This project will make advances in Language Technology to yield practical dialogue systems that support
realistic spoken interactions between people and computers. Key outcomes of the project are computational
models that: a) determine whether the computer is being spoken to; b) handle ungrammatical input, speech
disfluencies, speech recognition errors, and rambling utterances; c) interpret gesture and spoken language
in combination; d) combine outputs from different types of interpretation mechanisms; e) determine whether
there is a breakdown in communication; and f) generate responses. Outcomes (a-e) overcome current
problems in the interpretation of spoken utterances, and outcome (f) enables the system to contribute to the
dialogue.