A unified framework for symbol grounding in human-machine interactions

Abstract

We find the need to endow these machines with increasing degrees of ability to understand the human, so that they can dialogue with human. How symbols can acquire meaning for the system itself, independent of external interpretation, is an unsolved problem. After discussing some backgrounds, this paper introduced dynamic-programming reinforcement learning framework subject to Z-condition requirements, and it also represents a specific reality of the semiotic process in a multi-agent setup. This paper suggested that current research on symbolic grounding can be unified through reinforcement learning with multi-agent system settings.