2. GenAI models for ATM

Generative Artificial Intelligence (GenAI), has recently emerged as a powerful technology with diverse applications across multiple domains, including natural language processing. In this project, we explored the possible applications of GenAI in the field of air traffic management (ATM), aiming to find areas where GenAI can add value over current solutions. A literature review was conducted, concluding with a catalogue of candidate use cases. From this list a specific use case was selected and further developed as a proof of concept to demonstrate the impact of GenAI technologies. The selected use case is titled “Annotating NOTAMs with a Tag System using GenAI”, for which we developed a solution leveraging the bi-directional encoder representations from transformers (BERT) large language model (LLM) architecture. The problem was framed as a multi-label classification problem, allowing multiple tags per NOTAM, and a semi-supervised training scheme, thus leveraging unlabelled data to improve the generalizability and robustness of the model. We tested the model on a test dataset to measure the improvement in accuracy achieved by the proposed framework.