Aviation safety is paramount, demanding precise analysis of safety occurrences during different flight phases. This study employs Natural Language Processing (NLP) and Deep Learning models, including LSTM, CNN, Bidirectional LSTM (BLSTM), and simple Recurrent Neural Networks (sRNN), to classify flight phases in safety reports from the Australian Transport Safety Bureau (ATSB). The models exhibited high accuracy, precision, recall, and F1 scores, with LSTM achieving the highest performance of 87%, 88%, 87%, and 88%, respectively. This performance highlights their effectiveness in automating safety occurrence analysis. The integration of NLP and Deep Learning technologies promises transformative enhancements in aviation safety analysis, enabling targeted safety measures and streamlined report handling.


    Access

    Download


    Export, share and cite



    Title :

    Aviation Safety Enhancement via NLP & Deep Learning: Classifying Flight Phases in ATSB Safety Reports


    Contributors:

    Publication date :

    2025


    Remarks:

    NLP, Aviation Safety, ATSB, Deep learning, Flight phase. arXiv admin note: substantial text overlap with arXiv:2501.01694


    Type of media :

    Preprint


    Type of material :

    Electronic Resource


    Language :

    English







    Flight Safety in Business Aviation

    Robinson, J. B. / Royal Aeronautical Society | British Library Conference Proceedings | 1998


    Mining and Classifying Aviation Accident Reports

    Srinivasan, Prabhakar / Nagarajan, Venkataramana / Mahadevan, Sankaran | AIAA | 2019