Grammer correction issue

[Mansi9991]

I am transcribing an audio file using the small model.. Everything is perfect while transcribing the audio..
But my question is, The ' WHISPER' model does auto-correction in grammer..

Such my audio file contains :
"In this course, will we...."

But it returns:
"In this course. we will..."

The problem is :
It corrects the grammer mistakes while transcribing audio file

Here I am attaching the link to the audio file::
'https://drive.google.com/file/d/1lNjjKvZxJ8G_pH7i19oMhxLf2Oj5loiE/view?usp=share_link'

[phineas-pta]

with whisper the result is "we will", with wav2vec2 it's "will we", maybe the audio too noisy to decide

but if i read the whole transcript, "we will" is more gramatically correct

[Mansi9991]

Yes it is correct gramatically but audio file contains 'will we' so the result should be 'will we' but it returns 'we will'

[phineas-pta]

how could you be sure it's "will we" ? did you test with other models ? for me the audio is too noisy to hear properly

[Mansi9991]

Yes.... I have tried with other models also and also done after denoising the audio file but still no change...It gives 'will we' after I change the value of 'compression_ration_threshold' to '1.2' from '2.4' but it affects the other transcription in the audio file but it is still not every time I get 'will we' after changing the value

[phineas-pta]

sorry i can't help, some models i tried (paddlespeech, nemo stt_en_conformer_transducer_xlarge) return another text: "will be [exploring]", i think it should be audio quality problem, not sure of auto-correction