Knowledge on demand: the future of decision support systems?

Abstract

The availability and quality of the information system supporting crisis management depends on the existence of up-to-date, comprehensive and reliable knowledge bases, despite the multidisciplinary nature of the information to be exchanged, the actions, and the decisions to be taken. This article proposes a method to benefit from pre-trained language models to automatically update these knowledge bases at both design time and run time. We offer a model-based extraction process and a dedicated automated validation process. These two contributions are illustrated and discussed around a case study: an emergency decision support system used to prepare for and respond to an unexpected mass gathering at an abandoned French aerial base.