Using List of Values in Document AI to classify documents and other validation tasks

Document AI (also known as Document Information Extraction previously) helps you to process large number of business documents of various layouts.

With the advancement of Generative AI and LLM, Document AI can now process various document types and more complex documents. Thanks to the support for list of values data type, it’s more versatile than you might expect.

Let’s take a quick tour of what’s possible — and maybe spark some ideas for your next project.

Prerequisites:

Subscription to Document AI (Premium Edition)

Document Classification

Let’s say you’ve got a stack of documents — invoices, contracts, shipping receipts, and tax forms. Normally, sorting them would take hours.

With Document AI and natural language prompting, you can now say:

“Classify this document as one of: [‘CertificateOfCompliance’, ‘DeliveryOrder’, ServiceReport’]”

The clearer the prompts, the better the results.

The following shows an example of a schema that used list of values to classify documents.

You can add a total of 20 values here for this data type.

Extension

I applied Document AI’s list of values feature to detect the presence of signatures within documents. While signature detection wasn’t the primary objective, leveraging this approach helps automate a key part of document validation.

The example below uses list of values to determine if there’s a signature.

To bring everything together, I created a custom schema that performs two key tasks: detecting the presence of a signature and classifying the document type. For this demonstration, I used a sample Certificate of Compliance. The model successfully identified that a signature was present and classified the document as a Certificate of Compliance. The output below illustrates the results.

Now Supports More Languages!

A major upgrade to Document AI is its expanding multilingual support. Now, you can extract, classify, and analyze documents written not just in English, but also in languages like Malay, Spanish, French, Japanese, and more.

In this post, I introduced how to leverage Document AI’s list of values feature to enhance your projects. . While signature detection wasn’t the primary objective, this use case highlights the flexibility and potential of combining structured schemas with prompting. There’s plenty of room to expand on this approach for more advanced document processing tasks.

Document AI (also known as Document Information Extraction previously) helps you to process large number of business documents of various layouts.With the advancement of Generative AI and LLM, Document AI can now process various document types and more complex documents. Thanks to the support for list of values data type, it’s more versatile than you might expect.Let’s take a quick tour of what’s possible — and maybe spark some ideas for your next project.Prerequisites:Subscription to Document AI (Premium Edition) Document ClassificationLet’s say you’ve got a stack of documents — invoices, contracts, shipping receipts, and tax forms. Normally, sorting them would take hours.With Document AI and natural language prompting, you can now say:“Classify this document as one of: [‘CertificateOfCompliance’, ‘DeliveryOrder’, ServiceReport’]”The clearer the prompts, the better the results.The following shows an example of a schema that used list of values to classify documents. You can add a total of 20 values here for this data type. ExtensionI applied Document AI’s list of values feature to detect the presence of signatures within documents. While signature detection wasn’t the primary objective, leveraging this approach helps automate a key part of document validation.The example below uses list of values to determine if there’s a signature. To bring everything together, I created a custom schema that performs two key tasks: detecting the presence of a signature and classifying the document type. For this demonstration, I used a sample Certificate of Compliance. The model successfully identified that a signature was present and classified the document as a Certificate of Compliance. The output below illustrates the results. Now Supports More Languages!A major upgrade to Document AI is its expanding multilingual support. Now, you can extract, classify, and analyze documents written not just in English, but also in languages like Malay, Spanish, French, Japanese, and more. In this post, I introduced how to leverage Document AI’s list of values feature to enhance your projects. . While signature detection wasn’t the primary objective, this use case highlights the flexibility and potential of combining structured schemas with prompting. There’s plenty of room to expand on this approach for more advanced document processing tasks. See also:Tutorials:Setup BTP Trial AccountSetup Document AI in BTP Trial AccountList of values data typeSupported Languages Read More Technology Blog Posts by SAP articles

#SAP

#SAPTechnologyblog