How can I increase the chance to return text/words for all handwriting in scanned documents with mixed print text and handwriting?

Question

I used azure form recognizer api (non-custom model) for scanned document with mixed printed text and handwriting, sometimes, it missed to return handwritings - missed return text/words for some handwriting. also sometimes, for handwriting, it returned with style ='other' instead of 'handwriting'. How can I increase the chance to return text/words for all handwriting? Please help. Thank you!

Answer 1

Hello @ToSuccess

Thanks a lot for reaching out to us in Microsoft Q&A platform. I see you are working on Form Recognizer API, and you are looking for more accuracy to recognize mix content.

I am sorry there is no way to improve the chance in API since the model is prebuilt model and everything is fixed, it calls the computer vision API for analyzing.

Which prebuilt model and version you are working on? Could you please share the part of handwritten content which API recognized as other? I can forward it to product team as evidence to improve in the near future. If that's an obvious word which is handwritten but recognize as other, we can say that is a bug.

I hope this helps.

Regards,
Yutong

-Please kindly accept the answer if you feel helpful to support the community, thanks a lot.