Document Information Extraction - Template Inaccurate

[Vishal_97]

Hi All,

I want to use the templates feature of Document Information Extraction service for an custom invoice which has purchase order header and item details. The invoice looks like below screenshot.

The issue I am facing with a custom template is that, the line items are being read inaccurately. Even though I have annotated the documents properly for the position of line items and fields as below.

I have uploaded up to 4 documents with same annotations. Still the Extraction results comes like how its shown below.

The line items are ready fully inaccurately. Could anyone please help in addressing the issue here?

Thanks & Regards,

Vishal Raghav

[tobias_weller · Accepted solution]

Hi Vishal,

I just noticed that this question hasn't been answered, please apologize.

In order to analyze this, we need to look into your schema and the documents, therefore, it would be best if you create a support ticket (you should also be able to do this through our user interface).

One thing you can evaluate is whether the schema you are using has default extractors for the line items. If yes, you could evaluate whether removing all default extractors for all line item fields (for header fields you can keep them) improves the result.

Best regards,
Tobias

[kpmgbpa · Accepted solution]

Hello Tobias,
We are facing similar issues i.e. line items are highly inaccurate while using SAP Build Process Automation - Document Information Extraction Custom Templates. Despite of annotating correctly and inserting 2 invoices for a template the line items output is either absent or completely inaccurate for at least 5 out of 10 templates we are working with.

Also since we are using the trial account for SAP BPA, we were directed by the SAP Support to raise a Question herein on anwers.sap.com instead of an SR at me.sap.com.

Please help us to get in contact with the SAP Support Team directly to resolve these issues ASAP.

[tobias_weller · Accepted solution]

Hello,

Unfortunately, without seeing the samples and the schema definition it is difficult to provide any guidance and for sharing these files in a compliant way, we need to use the official support channel.

Other than that the main thing you can try is to remove default extractors on line items:

"One thing you can evaluate is whether the schema you are using has default extractors for the line items. If yes, you could evaluate whether removing all default extractors for all line item fields (for header fields you can keep them) improves the result."

Best regards,

Tobias