Azure AI Document Intelligence Code Samples

Prebuilt models allow you to process specific document types without training custom models. They are optimized for extracting specific fields from common business documents:

• Financial/Transaction: Invoice, Receipt, Pay stubs, Checks, Credit/Debit card.
• Identity/Legal: Identity, Contract, Marriage certificate, Health Insurance card.
• Mortgage: US Mortgage 1003, US Mortgage 1004, US Mortgage 1005, US Mortgage 1008, US Mortgage closing disclosure.
• Tax: US Tax W-2, US Tax 1040, US Tax 1095, US Tax 1098, US Tax 1099.

The prebuilt-idDocument model is used to extract information from various identity documents. The supported document types vary by region:

• Worldwide: Passport Book, Passport Card
• United States: Driver License, Identification Card, Residency Permit (Green card), Social Security Card, Military ID
• Europe: Driver License, Identification Card, Residency Permit
• India: Driver License, PAN Card, Aadhaar Card
• Canada: Driver License, Identification Card, Residency Permit (Maple Card)
• Australia: Driver License, Photo Card, Key-pass ID (including digital version)

The prebuilt-receipt model is used to extract structured information from various types of receipts. It supports different receipt categories including thermal receipts (retail, meals, parking), hotel receipts, credit card receipts, gas receipts, and parking receipts. Use the Model ID prebuilt-receipt when making API calls to trigger this specific extraction logic.

Model ID: prebuilt-receipt

The model extracts an array of objects under the key IncomeOrLossFromRentalRealEstatePropertyDetails. Each object in this array represents a specific property or royalty entry and contains the following nested fields:

• PhysicalAddress (address type)
• FairRentalDays (number)
• PersonalUseDays (number)
• QJV (boolean)
• Various box values (Box3 through Box22, all number type)

{
  "IncomeOrLossFromRentalRealEstatePropertyDetails": [
    {
      "PhysicalAddress": { "street": "...", "city": "..." },
      "FairRentalDays": 1,
      "PersonalUseDays": 364,
      "QJV": true,
      "Box3": 123456
    }
  ]
}

The Box17 field is an array of objects used to capture State Specific FUTA (Federal Unemployment Tax Act) details. Each object in the array contains the following properties:

• NameOfState (string): The name of the state.
• TaxableWage (number): The taxable wage for that state.
• StateExperienceRatePeriodFromDate (date): The start date of the state experience rate period.
• StateExperienceRatePeriodToDate (date): The end date of the state experience rate period.
• StateExperienceRate (number): The state experience rate.
• BoxE (number): Value from Box E.
• BoxF (number): Value from Box F.
• BoxG (number): Value from Box G.
• ContributionPaid (number): The contribution paid.

Azure Document Intelligence provides document analysis models to extract text and structure from forms. Use these models when you need to convert unstructured documents into structured, business-ready content.

• Layout: Use this to extract text, tables, and the overall document structure.
• Read: Use this specifically for extracting printed and handwritten text.

The prebuilt-contract model is a specialized Azure Document Intelligence model designed to extract key information from legal contracts. As of the 2024-11-30 (GA) release, it supports English (United States, en-US) and extracts structured fields including contract identifiers, parties, dates, and jurisdictions.

Model ID: prebuilt-contract

When prebuilt models do not meet your requirements, you can use custom models to handle specific document types or layouts:

• Custom template: Best for extracting data from documents with static, unchanging layouts.
• Custom neural: Best for extracting data from mixed-type documents where layouts may vary.
• Custom composed: Allows you to extract data using a collection of different models.
• Custom classifier: Used to identify and categorize document types (classes) before routing them to the appropriate extraction model.

Azure Document Intelligence provides a prebuilt model specifically for analyzing US W-2 tax forms. This model extracts key information such as employee and employer details, tax withholdings, and state/local tax information.

Model ID: prebuilt-tax.us.w2

Supported Languages and Locales:

• English (en-US)

The Azure Document Intelligence prebuilt model for US tax form 1099-B allows for automated extraction of tax year, payer, recipient, transaction summaries, and individual transaction details.

Model ID: prebuilt-tax.us.1099B

Supported Language/Locale: English (en-US)

Model ID: prebuilt-tax.us.1099B

Since document analysis is an asynchronous operation, you must use getLongRunningPoller to create a poller from the initial response. Use poller.pollUntilDone() to wait for the final result.

import {
  getLongRunningPoller,
  AnalyzeResultOperationOutput,
  isUnexpected,
} from "@azure-rest/ai-document-intelligence";

if (isUnexpected(initialResponse)) {
  throw initialResponse.body.error;
}
const poller = await getLongRunningPoller(client, initialResponse);
const result = (await poller.pollUntilDone()).body as AnalyzeResultOperationOutput;
console.log(result);