Form Recognizer Supervised Training Question

Question

I'm amazed by the result of MS Form Recognize and I think MS Form Recognizer has phenomenal potential. While giving a trial on a ballot form, I see the output is superb. However, i do find something I cannot do and I want to understand why.

To demonstrate the issue I am facing, i added a screenshot here below. For the checkbox marked as pink, Form Recognizer indicates it is "unselected", however, i want the recognition engine to think it is "selected", therefore, i would like to change the value from "unselected" to "selected" and expected that the supervised learning engine would take my specified "output" value as a guidance so that it can intelligently recognize this type of checkbox as 'selected" even though it is not properly marked as selected. Is this something that can be done or not possible to do? I just started looking into MS Cognitive Services and I have billions of ballot forms to process if i can make this work.

Your suggestion is very much appreciated.

Kevin

Answer 1

@shengk

Hello, hope you are doing well.

Since the result for checkbox selected or unselected is from OCR directly, I am sorry that this can not be trained at this moment for v2.1.

I have forwarded this feedback to Form Recognizer product group for future optimization.

Form Recognizer V3 is on preview now with form recognizer studio, I hope you enjoy this product. https://learn.microsoft.com/en-us/azure/applied-ai-services/form-recognizer/quickstarts/try-v3-form-recognizer-studio

Sorry for inconveniences.

------------------------------

• Please don't forget to click on or upvote button whenever the information provided helps you. Original posters help the community find answers faster by identifying the correct answer. Here is how
• Want a reminder to come back and check responses? Here is how to subscribe to a notification
• If you are interested in joining the VM program and help shape the future of Q&A: Here is how you can be part of Q&A Volunteer Moderators

Answer 2

Thank you Yutong for checking out the issues we're facing. I look forward to future improvement with the new version of recognizer.
I will certainly try the new version.

Thanks,
Kevin