Corpus-based planning of deictic gestures in COMIC

Abstract

We describe the recording and annotation of a corpus of role-playing dialogues in the domain of the COMIC multimodal dialogue system. We give some generalisations about the use of deictic gesture in this task, and show how those findings are currently used in the presentation-planning module of the dialogue system prototype. Finally, we describe how the use of gestures in the system will be evaluated, and outline the next steps in the development of the module. © Springer-Verlag 2004.