[Python] How to Extract Multiple Unlabeled Form IDs Using Azure Document Intelligence Neural...

I’ve created a Custom Extraction Model and trained the model using the Neural approach in Azure Document Intelligence (Form Recognizer). I’m successfully extracting most of the required fields, including coverage limits and Form IDs. However, I’m facing a challenge:

Form IDs in the document don’t have any associated label or key — they’re just raw values (e.g., BB 89 56 89 20, LM 20 68 90 32, CD 63 78 (FT. 0/ 17)) that appear in endorsement sections or at the bottom of the page. Currently I am able to fetch only single form id but i need all of them which are present in COI doc.

Is there a way to extract all such unlabeled Form IDs, ideally as multiple values for a field like form_ids[] using the neural model?

Appreciate any tips or strategies if anyone has dealt with similar use cases!

Continue reading...