For a company providing private aircraft charter services

The Problem

Every day, people send out hundreds of emails requesting private aircrafts for charter. On of our clients, True Aviation, wanted to connect people who had private aircrafts available along a specific route with people who wanted to travel on that route. For this, they required an interface that allowed email requests to be searchable by arrival and departure airports, dates & times, and aircraft types.

Our Solution

Since different people put in requests in different formats/writing styles, using a traditional rule-based parser was not a viable option. So, we instead opted for a machine learning-based approach. Our deep learning-based solution automatically extracts desirable information (departure, arrival airports etc.) from different requests and displays them in a terse searchable format.

One major challenge was that a single request could include multiple routes (e.g., the person wants to go from airport A to B on a specific date, and then from B to C on some other date). To solve this, our solution follows a two-stage strategy. In the first stage, it extracts the different airports, dates and times present in the request. In the next stage, it identifies which arrival airports, dates and times link to which departure airports.

Our final deliverable was a web-based interface which receives requests in real-time, processes them using our machine learning model and displays them.

Preview of the web application. The email from the customer is on the bottom right and the information extracted by the ML model is displayed in a concise format at the top

Results

The accuracy of our AI model for extracting data was around 98% and significantly enhanced operational efficiency, by saving dozens of hours spent each week by private aircraft charter agencies manually going through requests.