The field of aviation safety is increasingly leveraging Natural Language Processing (NLP) and Artificial Intelligence (AI) to enhance safety measures and operational efficiency. Recent developments focus on the application of deep learning models to classify and analyze safety occurrences from unstructured text narratives. These advancements aim to automate the analysis of safety reports, enabling the identification of critical safety issues and the classification of incidents by phase of flight or damage level. The integration of NLP and AI technologies promises transformative enhancements in aviation safety analysis, offering targeted safety measures and streamlined report handling. Challenges such as the need for large, annotated datasets and the interpretation of complex models are being addressed through innovative solutions like active learning for data annotation and explainable AI for model interpretation.

Noteworthy Papers

• Sequential Classification of Aviation Safety Occurrences with Natural Language Processing: Demonstrates the competitive performance of various deep learning models in classifying safety occurrences, with sRNN slightly outperforming others in recall and accuracy.
• Natural Language Processing and Deep Learning Models to Classify Phase of Flight in Aviation Safety Occurrences: Highlights the effectiveness of NLP and deep learning models in inferring flight phases from text narratives, with sRNN greatly outperforming ResNet.
• Aviation Safety Enhancement via NLP & Deep Learning: Classifying Flight Phases in ATSB Safety Reports: Shows the high performance of LSTM, CNN, BLSTM, and sRNN models in classifying flight phases, with LSTM achieving the highest performance metrics.
• Exploring Aviation Incident Narratives Using Topic Modeling and Clustering Techniques: Provides insights into incident narratives through advanced NLP techniques, with LDA performing best in identifying latent themes.
• Phase of Flight Classification in Aviation Safety using LSTM, GRU, and BiLSTM: A Case Study with ASN Dataset: Underscores the capacity of single and combined RNN-based models to classify the phase of flight from raw text narratives, showcasing the benefits of combining architectures.