Implementation of Embodied Cognition in Multi-agent Neurocognitive Architecture

Cognitive linguists assume that we use conceptualization to categorize the world around us. Concepts are units of knowledge that we keep in our brains to think. Theories of embodiment argue that concepts are based not only on language input but also on movement and perception that co-occur with it. It means that direct experience means a lot to our ability to learn new things and words. This congruency is well-seen in the process of first language acquisition. Nowadays being surrounded by devices with elements of artificial intelligent, it is obvious that the most convenient way to utilize them is via habitual conversations. To achieve this level of communication intelligent systems have to obtain not only a thorough vocabulary, but also to be able to understand language at a deeper level, at the semantic ones. Formal representation of semantics has been a hinder in a way to develop language understanding systems for decades. There were two main reasons: (1) the way semantics was obtain and presented was vague; (2) tools for formal representation of semantics were not effective. In this article we will try to give a short overview of the theories of embodied cognition, which as we assume give a reliable explanation of how semantics is built in our brains. While these theories explain the nature of semantics and conceptualization, multi-agent system can be a great tool to model these processes.