Enhancing Causality Detection in Nuclear Event Reports through a Hybrid Approach

Abstract

This article explores the application of a hybrid approach for enhancing causality detection in Nuclear Event Reports (NERs), addressing the complexities of incident analysis within the nuclear industry. Nuclear Event Reports document critical incidents and deviations, serving as pivotal resources for safety enhancement and regulatory compliance. Traditional methods for causality detection often face challenges in handling the nuanced and diverse textual data found in NERs. The proposed hybrid approach integrates rule-based systems with machine learning techniques, leveraging the strengths of each to achieve more accurate and comprehensive causality extraction. Key findings highlight significant improvements in accuracy and efficiency compared to conventional methods, demonstrating the framework's robustness across varied incident types. The implications for the nuclear industry include enhanced safety protocols, regulatory compliance, and operational efficiency through advanced incident analysis capabilities. This study underscores the transformative potential of hybrid frameworks in bolstering safety management practices within high-risk industries like nuclear power.