A rational cognitive architectural model of language generation

Abstract

In this paper, a prototype model demonstrating the rationality in humans is developed. For queries to be answered, the internal knowledge representation mechanism is understood to be in the form of inter-related clusters permanently stored in the long term memory (LTM). In response to a query involving a single predicate (simplest case), only one cluster along with its various attributes and values may be retrieved. In contrast, a complex query involving multiple predicates requires the retrieval of multiple related chunks, forming a network of knowledge, often referred to as a concept map, from which the response is generated. An internal process that happens is the matching of predicate of the question with the facts in the knowledge-base and retrieval of the chunk(s) with the corresponding Attribute-Value pairs, through a process akin to unification, a substitution procedure. However, an unanswered query due to the absence of a specific knowledge generates a negative response.