Action-Centered Communication with an Embedded Agent

In most cases natural language processing is seen as an isolated cognitive capability of a system. Language understanding is often restricted to the mapping of natural language expressions into an internal semantic representation, whereas language production takes an explicit semantic representation as input, from which a natural language utterance is generated. The approach presented in this paper considers the ability to process natural language as a distributed competence of an embedded artificial agent. The agent is able to perceive the environment with its’ sensors (vision, tactile, telemetric) and processes natural language directives. In order to describe the ongoing action it is able to produce natural language utterances.