First look at Document Intelligence Studio

Azure AI Document Intelligence (formerly known as Form Recognizer) is an online tool that automates document processing using advanced AI and machine learning. It extracts and analyzes data from various documents, offering OCR, and key-value pair extraction.

Document Intelligence Studio is a user-friendly interface for you to experiment with pre-trained Document Intelligence models or to create your own custom models. 

Link: Document Intelligence Studio

Requirements

To use the Document Intelligence Studio, you need to create a ‘Document Intelligence’ resource or an ‘Azure AI services multi-service account’. The latter allows you to use multiple kinds of AI Services using one endpoint including the ‘Document Intelligence’ service.

Document Intelligence Studio

When you open the portal and scroll down the screen you will see 3 different types of models.

Document analysis

These are the base models upon which other models are built.

  • Read: Extracts text content from documents, including both printed and handwritten text.
  • Layout: Extracts text, tables, checkboxes, and figures from documents preserving the layout of the document.
  • General Document: This model is deprecated from version (2023-10-31-preview).

Prebuilt models

The prebuilt models are pre-trained models like invoices, receipts, or id documents available to use without the need to train your own. 

Custom models

 Here you can create your own custom models. There are 2 different types of custom models available. 

  • Custom extraction model: This model extends the Layout model and allows you to train a model that extracts text, tables, checkboxes, and figures from documents. You need a minimum of 5 example documents to train this custom model.
  • Custom classification model: This model allows you to train a model to classify documents. You need a minimum of 5 example documents per category to train this custom model.

Try it out

You can try out the prebuild and base models directly from the portal. Click the try it out link on the model you are interested in to open the analysis page. Here you can analyze files and see the results.

  1. Select one of the provided examples or upload your own file.
  2. Click Run analysis button
  3. Review the results
The results are split in 3 tabs. Content contains a textual representation of the extracted values. Result will show the json value of the analysis and Code will give you examples of how to call the model using Python, C# and JavaScript.
Facebook
X
LinkedIn
WhatsApp
Email