Computational Linguistics Colloquium

Thursday, 12 December 2013, 16:15
Conference Room, Building C7.4

Towards a model for incremental grounding in spoken dialogue systems

By analyzing the user's utterance while it is still in progress, dialogue systems and virtual humans can provide more natural, human-like conversational behavior. This includes (non-verbal) listening behavior, backchannels, quick responses and interruptions and collaborative completions. I will present recent work at the Institute for Creative Technologies aimed at incremental understanding and dialogue management while someone is speaking. A particular focus will be on a model of grounding (reaching assumed mutual understanding of what is said) that is updated very rapidly and facilitates feedback decisions while listening. These models have been implemented within a multi-party virtual human negotiation applications. I will also present work on providing additional listener feedback to indicate attitudes toward the content as well as level of understanding.

If you would like to meet with the speaker, please contact Michael Wiegand.