How to retain interjections within the transcription?

[chuuck]

Hi there,

When testing the Whisper against real world scenarios, it seems like Whisper automatically removes or either doesn't recognise interjections in speech. With the samples I have played around with it seems to either skip interjections or make up a word that doesn't even exist (further digging showed me that those tokens normally have low confidence score). Some examples of interjections are in this website here.

My goal is to include them within the transcription. Do you know if there is a solution possible that would allow including interjections?

Thank you!

[jianfch]

See #286