Generation of inner and outer speech by means of situational context

Abstract

The work solves the question of establishing semantic and situational context in project ARS (Artificial Recognition System). Context is needed in ARS, if two agents have to solve a question in a certain problem domain. It is defined through a domain specific ontology and a general ontology. Talking about a distinctive problem requires agents to know the problem domain and to have specific knowledge about it. Solving a general problem requires less domain specific knowledge, but more abstract knowledge. For that reason, two different ways of knowledge bases exist. In order to develop situational context the agent retrieves memory traces from these ontologies and compares them to the actual situation. If the perceptions from the actual situation match with the previous stored situational memories, situational context is recognized. By establishing situational context, the agent is able to produce speech thought statements in a certain scenario. The situational context is triggering an action, such as speech for example. The result of the work is the generation of speech within a situational context. In order to produce speech a modular model from psycholinguistics has been integrated. The model is applied in a simplified form onto the ARS model to enable language ability (inner and outer speech) for the agents. The language is triggered through situation, which is by means of abstraction also able to generate speech in new situations, as it is looking for known structures in the old pattern. The abstraction mechanism is looking for matches in a new situation until something is found. If this is the case a known speech statement can be transfered as well to a new situation.