Confidence Threshold of document in Document Understanding

Hi All,

Currently implementing solution using Document Understanding with AI Center.
Need to implement logic where document with less than confidence threshold should go to manual queue.
We can get confidence of individual fields but how get Confidence threshold of entire document. or how to get confidence of entire document.


what are you facing issues with?

When you use the Export Extraction Result activity, that will give you a JSON that will have the confidence.

If you need to know the steps, this video might help. If you have done all this and have any specific doubt you may give us more information.
Confidence in Document Understanding

Hi @abhilash.bhanwal,

Can you please share which point you are trying to check confidence value for?

If trying to check after classification of documents, then here confidence level is checked at document level.

If trying to do same after extraction of data from document, then here confidence level is saved at fields level.


thanks Rahul for this video. I have already gone through it.
I have one more doubt after extraction using ML extractor and export extraction results is giving duplicate tablesimage

is it something in built or if not then whats the solution for it

Thanks Sonali, I have another issue for which i have replied. Please check if you can help on that

Hi @abhilash.bhanwal,

Entries are seen twice because when data tables are saved for extraction results, one would be simple fields and other would be simple fields - formatted.

Try saving extraction results in a csv file, you would notice 2 sheets being created for 1 doc with some differences in the format of the fields extracted.

Attaching an example as below from one of my DU runs:
invoice-02.xlsx (9.3 KB)