RecoMap - a semi-automated tool for analysing railway accident recommendations across jurisdictions and over time

Abstract

To maintain a safer railway operational environment, recommendations are proposed by independent investigators after accidents. Despite a considerable number of (sometimes similar) recommendations made across jurisdictions, practitioners suffer from a lack of synthesised recommendations made across jurisdictions and time due to the high complexity of analysing textual data. To fulfil the gap, an auto mated tool for the analysis of accident report recommendations is developed, allowing the railway industry to learn from other countries. The Structural Topic Model (STM) is used to extract critical insights from recommendations to depict how independent railway accident investigators mitigate risks observed. Empirical data is retrieved from official railway accident reports published by Rail Accident Investigation Branch (RAIB), Australian Transport Safety Bureau (ATSB), National Transportation Safety Board (NTSB) and Transportation Safety Board of Canada (TSB). The resulting RecoMap is developed as a framework to help practitioners learn across jurisdictions and time. The study also identifies a transition from making interfering recommendations addressing operational issues to making supportive recommendations addressing organisational issues in the railway industry across countries. Additionally, the concept of triple-loop learning is insufficient in the railway industry of the investigated jurisdictions, implying that current practices might result in railway accidents that could have been prevented by learning from other jurisdictions and implementing corresponding mitigation measures in advance.