What is it about?
This study investigates how Natural Language Processing (NLP) and Artificial Intelligence (AI) models can be used to classify safety occurrences in aviation by analyzing the text narratives of accident reports. The main goal is to determine how accurately the damage level caused to an aircraft can be predicted from these narratives. By testing a variety of deep learning models, including LSTM, BLSTM, GRU, and several combinations, the study demonstrates how well these models can process the unstructured text of safety reports and classify them into different damage levels.
Featured Image
Photo by Possessed Photography on Unsplash
Why is it important?
Aviation safety is crucial, and accurately interpreting incident reports is essential for preventing accidents and improving safety measures. The ability to automatically classify and categorize safety occurrences from text narratives can significantly help aviation professionals make better decisions based on data. The study's findings show that AI and NLP models can help process vast amounts of safety data quickly and accurately, ultimately enhancing decision-making and safety protocols in the aviation industry
Perspectives
The findings from this study highlight the power of NLP and deep learning models in transforming raw, unstructured text into meaningful safety insights. With accuracy levels exceeding 87.9%, the models demonstrated high performance across multiple evaluation metrics, including precision, recall, and F1 score. This suggests that AI can play a key role in improving safety in aviation by providing tools that allow stakeholders to better assess the severity of safety occurrences, and thus, take proactive measures to mitigate risk. The research also paves the way for further exploration into improving these models and making aviation even safer.
AZIIDA NANYONGA
University of New South Wales
Read the Original
This page is a summary of: Sequential Classification of Aviation Safety Occurrences with Natural Language Processing, June 2023, American Institute of Aeronautics and Astronautics (AIAA),
DOI: 10.2514/6.2023-4325.
You can read the full text:
Contributors
The following have contributed to this page
