Automated schema generation

Preview

This product or feature is subject to the "Pre-GA Offerings Terms" in the General Service Terms section of the Service Specific Terms. Pre-GA products and features are available "as is" and might have limited support. For more information, see the launch stage descriptions.

Document AI's automated schema generation lets you automatically generate a document's schema from a test document you supply. Then, you can approve or decline the schema and edit it manually. This saves time and effort when defining the document schema for your custom processor and lets you focus on refining the schema.

Automated schema generation also has a wider knowledge base on creating high quality schemas. This can potentially improve document extraction quality.

Generate a schema

If you haven't already snapshotted or fine-tuned your processor, take the following steps.

Note: If you've snapshotted or fine-tuned the processor, the option in the first step isn't displayed. The existing fields can't be deleted.

1. On the Get started tab, select Generate schema from document. The Edit schema page opens.

   

2. Use the input field to select a local file file to generate the schema, or use the Browse option to upload a sample document.

   

   Tip: Enter a prompt in Generate a schema prompt to provide additional input and direction for the generator.

3. Select Generate Schema.

4. Review the generated schema preview. You can choose accept or reject the schema with Apply schema or Abort schema, to try again with a different sample document or prompt.

   

   Tip: You can use the View More option by each field to edit the auto-generated schema after applying it.

5. Optional: If the resulting schema doesn't serve your needs, or requires more than manual editing, you can enter a prompt in Generate a schema prompt to provide additional input and direction for the generator. Try prompts such as: specifying the document type, stating the most important parts of the document, or suggesting a target number of entities.

What's next

• Learn about uptraining a specialized processor.
• Explore derived signature detection.