Adaptive behaviors of intelligent agents based on neural semantic knowledge

This paper presents a new approach that combines the neural network technology with the semantic network-based knowledge representation in realizing the intelligent, emotional, adaptive and believable behaviors of consulting personal agents. This knowledge representation framework is deployed in the CyMON agent platform to develop intelligent assistant, Consultant and marketing agents for a variety of application areas. In this context, semantic network-based knowledge representation is regarded as a natural way to simulate the human conception of world and entities, resulting in intuitive and explicit representation of concepts and relations that can be subject of dynamic agent learning. Neural network technology is used to simulate a belief network which reflects the dynamic adaptive agent behaviors. The network-oriented representation of the knowledge enables the relatively straightforward integration of semantic and neural network technologies.