Integrating natural language generation and model-based reasoning for explanation generation

Natural Language Generation (NLG) systems can utilise external reasoning to explain events within a given domain. Most often, these explanations are derived from compiled knowledge such as that found in expert systems, which can lead to gaps in the explanation system's ability to justify the outcome of the reasoner. We propose the use of a deeper knowledge source for explanation generation and show how to utilise a model-based approach that uses first-principles knowledge of the problem domain to generate and justify more in-depth explanations of events in the domain. Our integrated framework of NLG and model-based reasoning makes two contributions: the articulation of model-based reasoning and justification in natural language, and the natural language generation engine exploiting the deeper knowledge available in the underlying model.