Assuming 2-3 forms can come clubbed together in one pdf file and order is dynamic.
How would one go to develop the solution.
Note-tested individually all forms are getting classified and extracted correctly.
Hi @Mudit_Choudhary ,
Since each form is classified and extracted correctly on its own, the main challenge is handling multiple forms in a single PDF.
A good approach is:
- Use Document Separation/Splitting first to identify where each form starts and ends.
- Classify each document segment independently using the Classifier.
- Route each classified section to its corresponding extractor/model.
- Merge the extraction results into a structured output if needed.
- If the order is dynamic, do not rely on page numbers. Let classification determine the document type for each separated section.
- Add validation for cases where the same form type appears multiple times in one PDF.
Thanks
Hi @Mudit_Choudhary,
Refer to this documentation. This model has the document splitting capability, if there are multiple forms within a single document.
Cheers!
If multiple forms can be in one PDF and the order is dynamic, use Document Understanding classification to identify each form first, then route each classified form to its respective extraction model.
Since individual forms are already working correctly, the main focus would be classification → identify form type → extract. Avoid relying on fixed page order.
Hi @Mudit_Choudhary, since each form already classifies fine, what’s missing is splitting. In your modern project, turn on the new splitter and classifier model with splitting enabled, then train on a few real combined PDFs and mark where each form starts. At runtime you get one classification result per form, so loop over them and extract each one in whatever order they come. It’s in Preview and only available for EU and US tenants.