| Towards Customizable Individualized Dialogue Systems (2008) | |||||||||||||||
Abstract | |||||||||||||||
| Spoken language dialogue systems are one of the few extant examples of an intelligent artifact that can interact with humans. As such, they provide a platform for exploring fundamental cognitive science questions about the nature of human language processing and interaction. Recently, research in this area has moved towards using empirical natural scientific methods that rely on the generation of theoretically motivated hypotheses, derived from cognitive and computer science, which are then tested via experiments on real-time systems with real users. One of the most consistent findings from research on the modeling of human cognition is individual differences in cognitive processing and style, which is as yet unaddressed by these systems. It is our aim to address that here. Research Problem: Current dialogue techniques are difficult to customize to new domains and to individual users, who may have quite different needs and preferences (e.g. expert vs. novice users, the elderly, business people vs. teenagers). In previous work, we’ve developed novel techniques for automatically training dialogue system modules for new application domains. These techniques suggest a new approach to building dialogue systems [Young02], which should allow systems to be automatically customized to individual users. However this promise of individualized systems has yet to be tested. | |||||||||||||||
Publication details | |||||||||||||||
| |||||||||||||||