Overview of Representation Languages for ECAs

Agents may be decomposed roughly into two main parts: the body and the mind. The body performs the animation. It may be 2D, 3D, cartoon-like, realistic... It may be based on MPEG-4 standard; it may follow H-Anim specification; its animation may come from motion capture, dynamic system or key-frames. The mind is responsible for the reasoning, the planning, the perception of the world it is placed in as well as of the user(s) or agent(s) it is talking to. The mind has to perceive and understand what is being said and what are the events happening in the world. It has to generate what to say and how to say it; it should also know how to react to the events, to trigger emotions or actions to undertake... Moreover, the mind reflects the personality and the identity factors of the agent. This report does not deal either with the body part, or with the mind part; it does not deal either from the input side (perception and understanding) or from the output side (generation and animation). This report aims at making the link between the body and the mind of the agent; this link may be achieved by developing a representation language that specifies the agent’s behaviours. Such a language serves as interface between the different modules of the architecture of an agent system. Moreover, the languages are used to control the animation of agents as well as they are used as a synchronization scheme between modalities. The languages may embed various levels of abstraction: ranging from the description of the signals (a smile, a head nod) to semantic information (rheme / theme, iconic), going through communicative function (performative, emotion) and syntactic function (question, statement).