Document Information Extraction

[former_member1317659]

Good Morning,

I am currently working with the Data Information Extraction SAP Service and I have some questions about it:
- Is it a Beta Version or was it released to customers?
- Is it possible to increase the number of templates in order to increase the number of documents from which it is allowed to correctly extract data? At the moment the only document with which the service works correctly is the invoice.
- Is it possible to use the Machine Learning in the Data Extraction Data? If yes, how is it possible to teach the service which are information to be extracted? Do we need to manually insert more and more times the same field and then the service learns it or there is another way for teaching it?

Thank you for your help.

Best Regards,

Lorenzo Storella

[tomasz_janasz · Accepted solution]

Hi Lorenzo,

let me answer your queries:

1. Document Information Extraction is generally available. Currently there is no Beta program in place. You can start using the service on BTP trial. PLease follow this tutorial: https://developers.sap.com/mission.cp-aibus-extract-document-service.html

2. Soon we will release Document Information Extraction, extension capabilities. The first capability will be the possibility to set up Templates and custom schemas.

3. Currently there is no automatic "feedback/learning" loop based on user corrections. The global models are managed and retrained by SAP.

Best regards,

Tomasz (from the PM team)

[former_member1317659 · Accepted solution]

Hi Tomasz,
thank you for your answer!
What do you mean when you say "Soon we will release..."? Which period of time are you talking about?
Thank you!
Lorenzo Storella

[Colin_Brain · Accepted solution]

Hi Tomasz,

Have I understood correctly that if the current models don't read the invoice correctly, we will have to correct the data every time - that there is no way to get the service to not make the same error over and over?

Regards

Colin

[tomasz_janasz · Accepted solution]

Hi Colin,

I think you are referring to so called "automatic retraining loop" or "continual learning". DOX service is not yet enabled with that. The machine learning models are maintained by SAP. These models can be further optimized. If you face issues with the service please open a support ticket, provide the description and the corresponding document sample.

https://launchpad.support.sap.com/

Component to be selected: CA-ML-BDP

Best regards,

Tomasz