Generating Natural Language Explanations from Knowledge-Based Systems Results, Using Ontology and Discourses Patterns

The understanding of results of Knowledge-based systems (KBS) working on complex Dynamic Systems (DS) requires expert knowledge and interpretation capability in order to make a correct analysis of observations at multiple scales and instants. Normally, these kinds of KBS generate extensive inference-trees before showing a definitive result to final users; these inference-trees are not included in the KBS outputs, but they could provide additional information to understand the functioning of the KBS, and also to understand the overall performance of a DS. This document describes a method to generate natural language explanations, based on the results reached by a KBS in respect to a DS behavior, using a specific ontology and discourse patterns. The input of the method is an intermediate-state tree (the inference-tree) and specific domain knowledge represented on domain ontology. The document describes also the software architecture to generate the explanations and the test cases designed to validate the results in a specific domain.