Verification of Safety Rules using NLP

A key step in the design of digitally enabled safety systems is the development of an Enterprise Architecture model (EA). The design of EA models tends to be a complex job that is usually performed by IT specialists that are not trained in safety. Very often, these EA models contain safety rules that are not well understood by IT specialists but they are of key importance for the safe implementation of the digital solutions. As part of safety directives, safety experts have to verify whether there are any safety issues in the EA model. This particular aspect of enterprise architecture verification is a manual process that is laborious and prone to error. This work investigates whether standard Natural Language Processing techniques (NLP) can help in the verification of safety rules within an enterprise architecture model. This paper demonstrates that this kind of verification is potentially very powerful but cannot be used on its own; ontological taxonomies are probably required as well.