Document Understanding : How to Extract Invoice Fields not Supported by Builtin ML Model

Hi Community!

I am currently working on a project to automate invoices via UiPath Document Understanding. I am able to extract fields through ML Extractor provided by UiPath. The supported fields are:

  • name
  • vendor address
  • billing name
  • billing address
  • shipping address
  • invoice number
  • po number
  • vendor VAT number
  • date
  • tax
  • total
  • payment terms
  • net amount
  • due date
  • discount
  • shipping charges
  • payment address
  • line item Description
  • line item quantity
  • line item unit price
  • line item amount
  • line item PO number
  • line number
  • line item part number
  • billing VAT number
  • currency

However, I am looking to be able to extract more fields from an invoice such as:

  • LineItem Unit Measure
  • LineItem Product Code
  • Subtotal amount.
  • Reference or order number.
  • Period of Performance, the Subscription period, expiring time/date.
  • Delivery/Shipping Date.
  • Discounts percentage
  • Early payment discounts options.
  • Advance Payments/Credits.
  • Late Payment fee or surcharge.
  • Balance Due.
  • Payment options/Method: Such as Credit card, Check, Bank transfer.
  • Banking information:-
  • Account Number.
  • Routing number.
  • Bank Name.
  • Bank address.
  • Payment instructions.
  • Contact Person/Contact
  • LineItem line tax amount
  • LineItem line tax rate
  • Invoice Notes

My questions are:

  1. Can I extend the existing ML model to support these fields?
  2. Do I have to create my own custom ML Model to support these fields?

Thanks in Advance!
Tooba Faroki

1 Like

I have the same question, is there any way to extend the extraction fields of the pre-built ML engines?

It is possible to train your own custom ML model using AICenter.

Is there any way to add these (above mentioned) fields in the prebuilt model provided by UiPath?

The pretrained model is designed to be extensible and retrainable by our users. We will take your asks into consideration as we look at possible future model enhancements.

1 Like

Thank you for your response! I am now trying to use data manager to upload data to AI Centre to close the loop.

Does this mean that after training the existing ML model, I will be able to extract the new fields mentioned in my original post?

Yes, you will need to select your new ML Skill from the ML Extractor itself and refresh the schema in the extractor configuration screen, but after that, you should see the new fields you have trained available in the drop-downs.

1 Like

I appreciate that!

This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.