Automated Aviation Occurrences Categorization

Abstract

Information about aviation events is collected by all participants in the aviation system, e.g. airlines, maintenance organizations, and air traffic controllers. Reporting and initial assessment usually involves assigning categories from a predefined nomenclature (scheme) aligned with the purpose of the reporting system and the established processing practices. Such manual categorization is time and resource consuming and, more importantly, limiting the application of the dataset. We apply and evaluate the effectiveness of a state of the art Neural Networks based algorithm for Natural Language Processing for classification of aviation safety report narratives. Multi-class, multi-label supervised learning is performed on two small datasets, 4500 and 8000 cases with 16 and 54 classes respectively, both extracted from the NASA Aviation Safety Reporting System. The results are promising if compared to recent studies and considering that an off the shelf algorithm without much customization is applied. Automatic categorizations can relief the current burden for manual categorization of the events by reducing the number of likely categories, targeting quality checks to most ambiguous records and applying new or updated classification schemes.