ANGELICA Choice of output modality in an embodied agent

1 Abstract The ANGELICA project addresses the problem of modality choice in information presentation by embodied, human like agents. The output modalities available to such agents include both language and various nonverba l signals such as pointing and gesturing. For each piece of information to be presented by the agent it must be decided whether it should be expressed using language, a nonverbal signal, or both. In the ANGELICA project a model of the different factors inf luencing this choice will be developed and integrated in a natural language generation system. The application domain is the presentation of route descriptions by an embodied agent in a 3D environment. Evaluation and testing form an integral part of the pr oject. In particular, we will investigate the effect of different modality choices on the effectiveness and naturalness of the generated presentations and on the user's perception of the agent's personality.