From negative statements to positive safety

The GB railway industry collects hundreds of close call records each day from workers. The information is provided as free-text descriptions of hazards on the railway. Frequently the hazards occur as a result of risk controls breaking down. Information on failed risk controls is an essential input to the railway’s safety management system, but can be lost in the huge volume of information being reported.
This paper introduces a method based on natural language processing techniques that can automatically extract information on failed risk controls from text-based close call records. The method could substantially reduce the amount of effort needed to obtain critical safety information and is presented in a form that could be used immediately for supporting safety management of the railway.