The triple model: Combining cutting-edge web technologies with a cognitive model in an ECA

This paper introduces a new model which is intended to combine the power of a connectionist engine based on fast matrix calculation, RDF based memory and inference mechanisms, and affective computing in a hybrid cognitive model. The model is called triple and has the following three parts: a reasoning module making advantage of RDF based long-term memory by performing fast inferences, a fast connectionist mapping engine, which can establish relevance and similarities, including analogies, and an emotional module which modulates the functioning of the model. The reasoning engine synchronizes the connectionist and the emotional modules which run in parallel, and controls the communication with the user, retrieval from memory, transfer of knowledge, and action execution. The most important cognitive aspects of the model are context sensitivity, specific experiential episodic knowledge and learning. At the same time, the model provides mechanisms of selective attention and action based on anticipation by analogy. The inference and the connectionist modules can be optimized for high performance and thus ensure the real-time usage of the model in agent platforms supporting embodied conversational agents.