Document Automation with prebuilt model

[Iro_]

Hello,

Our first impressions of the Document Automation solution are excellent. The prebuilt model gives accurate results so far. We are going to use it with invoices and purchase orders. The challenge is that we absolutely need at least one more field to be extracted from the document (IBAN - and maybe VAT number as well). What is the proposed solution for our situation? Is there any way to train the prebuilt model so that it can detect more fields? Should we extract this information through other ways (ocr raw text etc)?

Thank you in advance.

[CedrickB · Accepted solution]

Hi,

Glad to hear that you are satisfied with our solution.

We are going to enable customization of the Invoice prebuilt next year.

In the meantime, you have several options which have different cost implication

1. Create a fully customized Form processing model to get all fields. Impact: human effort to build and maintain the model

2. Create a custom Form processing model to perform the last mile. Call the prebuilt and then call the custom model to get the IBAN/VAT. Impact: the computing cost as each document must be analyzed twice.

3. Use Text recognition (Basic OCR) and perform some post processing rules for the missing field. We have provided a template

Impact: this model is cheap so even if you call the Invoice Prebuilt + Text Recognition for each document, the overhead of cost will be non significant. Though, there will be an implementation cost as finding the right post processing rules to capture the IBAN may be challeging. 

Let me know if this helps.

[Iro_]

@CedrickB Thank you for the thorough response!

I need a few clarifications.

In case we proceed with the third option (text recognition): Is it possible to add new fields (iban, vat number) to the existing table (Document Automation Data table) in order to show these fields to the user for validation via Document Automation Application? Or do we need to create a new table for these new fields? 

In general, are the existing tables customizable? For example, I don't see a relationship between Document Automation Data table (which is supposed to be the Header information of an invoice) and the Document Automation Table Data table (Line items information) and I am wondering, can we add this relationship or do we have to create new tables and relationships since this is a managed solution?

Thank you in advance

[CedrickB]

Hi,

Sorry for the late answer.

Document Automation solution doesn't support yet the Text Recognition model.

What you can do to have it show up in the Application, you could perform some hacks:

1. Select the Invoice Processing model in the "Document Automation application"

2. Go to Document Automation Taxonomy and rename the "Name" on one of the rows containing unused fields to "IBAN" and "VAT number".

3. Update the "Document Automation Processor" flow to call the Text Recognition to get the missing field and update the corresponding Datax (X corresponding to the field you have renamed in the "Document Automation table Taxonomy")

Let me know if this works.

If not we can have a call together so I could assist you.

[Iro_]

Thank you.