Is it possible to recognize non-verbal expressions or to customize the tool (Google Speech / IBM Watson) for this? Non-verbal expressions are pauses during speech, for example:
"hum... i would like to know hum... how do i connect YouTube to Google AdSense"
In the tests I have done so far this type of expression is ignored in the transcript
The IBM Watson Speech to Text service rolls up these as %HESITATION. if you are not seeing this then you might have the smart formatting option on, which you will need to switch off.
https://cloud.ibm.com/docs/speech-to-text?topic=speech-to-text-basic-response#hesitation