I got the following coding question for a Machine Learning engineer with Axon. This is the copy/pasted question description. The interviewer afforded me about 22 minutes to solve this, which I was not able to do successfully.
You have an array of (audio) frames representing an audio file,
and a text file with the corresponding transcription.
Can you segment the audio into sets of monologues for each speaker?
The transcription file follows the format:
TIMESTAMP(S)\tSpeaker\tText
FPS: 20I asked for an example output, and the interviewer provided me this, verbatim:
Input
audio_frames = [f1, f2, f3]
transcription = [10\tJohn\t'Hi Kate', 12\tKate\t'Hi John', ...]
Output
"John" : [f1, ... f11], [f23, ... f25],
"Kate" : ...Comments (5)