AvNav has built a next-generation neural network model to transcribe ATC audio in real time. We demonstrated it at Sun ’n Fun, and the response was very encouraging.
Pilots generally liked the idea of ATC text being shown next to aircraft on a 3D map, as well as being able to look back through the transcript history if they missed something that may have been meant for them. Please see the screenshot below.
We are now looking for help improving the accuracy of the model using a human-in-the-loop review process.
If you are an experienced pilot, CFI, controller, airline pilot, or simply very comfortable with ATC phraseology, we could use help reviewing and correcting ATC audio transcripts. The work is remote, flexible, and paid. For qualified reviewers, we will pay at rates comparable to top CFI rates.
The goal is to improve the accuracy of a model we have already built. You do **not** need to transcribe audio from scratch. Instead, you would review model-generated transcripts, find the one or two words that are not quite right, and correct them using a simple web-based tool.
If interested, please DM me directly with a short note about your aviation/radio background.
Thanks,
Sanjay
AvNav
ps: Mods: if this belongs somewhere else, please feel free to move it.
Pilots generally liked the idea of ATC text being shown next to aircraft on a 3D map, as well as being able to look back through the transcript history if they missed something that may have been meant for them. Please see the screenshot below.
We are now looking for help improving the accuracy of the model using a human-in-the-loop review process.
If you are an experienced pilot, CFI, controller, airline pilot, or simply very comfortable with ATC phraseology, we could use help reviewing and correcting ATC audio transcripts. The work is remote, flexible, and paid. For qualified reviewers, we will pay at rates comparable to top CFI rates.
The goal is to improve the accuracy of a model we have already built. You do **not** need to transcribe audio from scratch. Instead, you would review model-generated transcripts, find the one or two words that are not quite right, and correct them using a simple web-based tool.
If interested, please DM me directly with a short note about your aviation/radio background.
Thanks,
Sanjay
AvNav
ps: Mods: if this belongs somewhere else, please feel free to move it.