Detecting inconsistencies of safety artifacts with Natural Language Processing Bachelor of Science Thesis

Abstract

This paper investigates a method that helps detect inconsistencies between safety-critical systems’ textual safety artifacts that safety cases rely on by involving NLP techniques. A design science research study was conducted in three iterations. I evaluate the method by conducting different experiments. The designed artifact identifies inconsistencies between different texts that are connected by trace links by checking similarities and word vectors of the texts. The results indicate that involving NLP technique word embedding can help with consistency classification. In conclusion, NLP techniques may help detect inconsistencies between safety artifacts that safety cases rely on, which is helpful to reduce the risk of system failures.