Multimodal Referential Acts in a Dialogue Game : from empirical investigations to algorithms

Abstract

Human-human conversation has often been heralded as a model for human-computer interaction, the rationale being that human-computer interaction can be improved significantly if it relies on those skills and abilities which come most natural to humans. The most rigorous application of this idea can be found in recent work on embodied conversational agents (ECAs; e.g., Cassell et al., 2000).