The Design and Implementation of SAMIR

摘要

This paper describes the architecture and the implementation of the SAMIR (Scenographic Agents Mimic Intelligent Reasoning) system, designed to implement 3D conversational agents. SAMIR uses an XCS classifier system to learn natural facial postures to be shown by the agents, during their dialogues with users. An administration program can be run by agent administrators in order to fine-tune the agent knowledge base and emotive responses. It results in a flexible approach to dialogue and facial expressions management, fitting the needs of different web applications such as e-shops or digital libraries.