Generation and Explanation: Building an Explanation Facility for the Explainable Expert Systems Framework

Generating explanations for expert systems has not been seen as a sophisticated generation problem in the past, and researchers working on expert system explanations (mainly researchers working on expert systems themselves) have been largely separated from the natural language generation community. In this paper, we argue that explanation for expert systems can benefit from the more sophisticated generation techniques being developed in computational linguistics and that explanation for expert systems actually provides a rich domain in which to study natural language generation. We describe our efforts to build a generation facility for the Explainable Expert Systems (EES) framework, presenting the requirements for this generation task and the issues addressed. We initially tried to use known natural language generation techniques but were led to design a new text planner, as these techniques did not fit our needs. This paper thus presents an overview of the generation facility being built for EES, including an ‘historical’ perspective that explains the decisions we made. Finally, we briefly present directions for future research.