Classify an Aeronautical Message (NOTAM) Using OpenAI ChatGPT

Classification of human-written messages into predefined categories is a problem every programmer or data scientist faces at some point in their career. These tasks are typically dealt with by extracting keywords out of a sample of messages such as social media posts or titles of books, and then mapping such keywords to categories. With the arrival of ChatGPT, these tasks can be approached in a totally different manner.

ChatGPT can be used as an expert, meaning we can give ChatGPT the created categories and then provide ChatGPT with a message and ask which of the categories would best fit this message. Finally, we can ask ChatGPT to structure the response in a structured format such as JSON. All of this can be done using a Python function.

In this project, we’ll explore such an example using Notices to Airmen (NOTAMs) messages issued by air traffic controllers to pilots as a source. The same approach can be taken on any other human-created message.