Those artifacts in the audio file should be trivial to correct, I suspect even a super simple outlier detection script would have worked. Should have been much

Those artifacts in the audio file should be trivial to correct, I suspect even a super simple outlier detection script would have worked. Should have been much easier than optimising the OCR (not to disparage the effort).