Unleash the Potential of Intelligent Document Processing using Azure Form Recognizer Studio

Intelligent Document Processing (IDP)

Intelligent Document Processing is one of the niche technologies which combines Artificial Intelligence (AI) and Machine Learning (ML) that automates and extracts information from unstructured documents like Invoices, Receipts, Business Cards, Identity Documents, Health Insurance Cards, and US Tax W-2.

US Tax 1098, US Tax 1098 - E, and US Tax 1098 - T are in private preview. User has to request if they want to access it and other benefits.

Azure Form Recognizer Studio is one of the Applied AI services and Cognitive Services which uses Optical Character Recognition (OCR) and Machine Learning (ML) technology to identify and extract key/value pairs and table data from form documents using prebuilt models or create your custom models.

Prebuilt Models

Extracting data and information from unique documents using Prebuilt Models.

• Invoices - Extract invoice ID, customer details, vendor details, ship to, bill to, total tax, subtotal, and so on.
• Receipts - Extract time, date of the transaction, merchant information, amount of taxes, totals, and more.
• Business Cards - Extract person's name, job title, address, email, company, and phone numbers from business cards.
• Identity Documents - Extract name, expiration date, machine readable zones, and more from passports and ID cards.
• Health Insurance Cards - Extract insurer, member, prescription, group number, and more information from US health insurance cards.
• US Tax W-2 - Extract employee, employer, and wage information from US W-2 Tax Form.
• US Tax 1098 - Extract mortgage interest information from US 1098 Tax Form.
• US Tax 1098 -E - Extract student loan information from US 1098 – E Tax Form.
• US Tax 1098 -T - Extract tuition information from US Tax 1098 – T form.
• Contracts - Extract the title and signatory parties' information, including names, reference names, and addresses from contracts.

Custom Models

Train custom models to classify documents and extract text, structure, and fields from your forms or documents.

Custom extraction model – Extract specific schema from your forms and documents.

Custom classification model – Split and classify documents.

Getting Started from Azure Form Recognizer

• Log into Azure Portal
• Search for the Form Recognizer service in the search pane.
• First, provide the Azure subscription
• Create a new resource group.
• Choose the Azure region and provide a Name.
• And then choose the pricing tier.
• Finally, click the Review + Create button.
• You will get a popup message stating that Validation passed.
• Then you click create button.
• The deployment started initialized in a minute or two it will become successful.

Once deployment succeeds, you will get the message "Your deployment is complete".

Creating Custom extraction model using Azure Form Recognizer Studio

• Click create new link.
• Then choose Template models.
• Create a new project.
  
• Click the continue button.
  
• Click the continue button.
  
• Click the continue button.
  
• Click Create Project button.

Steps for Creating Custom Extraction Model

It involves different phases or steps for creating a custom extraction model.

Phase 1: Prepare

In the prepare phase, the minimum number of documents is 5 to train the custom extraction model, and all 5 documents are in the BLOB container.

Phase 2: Label

In the label phase, you must label the documents with valid field names for training.

Click Train button

Click the Train button.

Phase 3: Model

In the Models phase, Model ID and status are generated.

Template - It allows Structured documents only, training time takes 1 to 5 minutes, and Multiple Languages support.

Neural - It allows Structured, Semi-Structured, and Unstructured documents; training time takes 20 minutes to 1 hour. Supports English and Preview support for French, Spanish, Dutch, German, and Italian.

Phase 4: Test

In the Test phase, I uploaded a sample invoice document and clicked Analyze button. It will automatically analyze form fields like Company Name, Charges, and Invoice numbers on the right side.

Conclusion

In this article, we created and successfully learned the implementation of Azure Recognizer Studio and its capabilities. How Intelligent Document Processing (IDP) is helpful for extracting data and insights from the real world, which helps to solve manual data entry problems for invoice document processing. Using Azure Form Recognizer Studio feature, which automates manual data entries and will be able to drive business values and makes the business a better impactful and impeccable solution.