ECAI 2004 Conference Paper

[PDF] [full paper] [prev] [tofc] [next]

Embodied Conversational Agent and Influences : Towards individualized agents

Vincent Maya, Myriam Lamolle, Catherine Pelachaud

In view of creating Embodied conversational agent (ECA) able to display different behaviors depending on factors such as context environment, personality and culture, we propose a taxonomy and a computational model of the influences these factors may induce. Influences act not only on the type of the signals an agent conveys but also on the expressivity of the signals. Thus, to individualize ECAs, we consider not only the influences acting on the agent but also the notion of expressivity. Expressivities arise at various levels of the agent representation (meaning, signal, behavior). In our ECAs architecture, the system takes an input text annotated with multi-modal communicative information tags. Considering the different expressivities representing the agent and the influences, the system instantiates the communicative tags into a sequence of \emph{expressive} signals. So for a same input text, our system associates to a given meaning different signals across modalities (face, gesture, posture, gaze), depending on agents' characteristics. In this paper, we also introduce a model of signals classification for a given meaning, based on expressivity, and a model of signals modulation.

Keywords: autonomous agent

Citation: Vincent Maya, Myriam Lamolle, Catherine Pelachaud: Embodied Conversational Agent and Influences : Towards individualized agents. In R.López de Mántaras and L.Saitta (eds.): ECAI2004, Proceedings of the 16th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2004, pp.1057-1058.

[prev] [tofc] [next]

ECAI-2004 is organised by the European Coordinating Committee for Artificial Intelligence (ECCAI) and hosted by the Universitat Politècnica de València on behalf of Asociación Española de Inteligencia Artificial (AEPIA) and Associació Catalana d'Intel-ligència Artificial (ACIA).