Document processing models

This content applies to: v3.1 (GA) | Previous versions: v3.0 v2.1

This content applies to: v3.0 (GA) | Latest versions: v3.1 | Previous version: v2.1

This content applies to: v2.1

Azure AI Document Intelligence supports a wide variety of models that enable you to add intelligent document processing to your apps and flows. You can use a prebuilt domain-specific model or train a custom model tailored to your specific business need and use cases. Document Intelligence can be used with the REST API or Python, C#, Java, and JavaScript client libraries.

Bounding box and polygon coordinates

A bounding box (polygon in v3.0 and later versions) is an abstract rectangle that surrounds text elements in a document used as a reference point for object detection.

• The bounding box specifies position by using an x and y coordinate plane presented in an array of four numerical pairs. Each pair represents a corner of the box in the following order: upper left, upper right, lower right, lower left.

• Image coordinates are presented in pixels. For a PDF, coordinates are presented in inches.

For all models, except Business card model, Document Intelligence now supports add-on capabilities to allow for more sophisticated analysis. These optional capabilities can be enabled and disabled depending on the scenario of the document extraction. There are seven add-on capabilities available for the 2023-07-31 (GA) and later API version:

• ocrHighResolution
• formulas
• styleFont
• barcodes
• languages
• keyValuePairs (2024-02-29-preview, 2023-10-31-preview)
• queryFields (2024-02-29-preview, 2023-10-31-preview) Not available with the US.Tax models
• searchablePDF (2024-07-31-preview) Only available for Read Model

Language support

The deep-learning-based universal models in Document Intelligence support many languages that can extract multilingual text from your images and documents, including text lines with mixed languages. Language support varies by Document Intelligence service functionality. For a complete list, see the following articles:

• Language support: document analysis models
• Language support: prebuilt models
• Language support: custom models

Regional availability

Document Intelligence is generally available in many of the Azure infrastructure regions.

Model details

This section describes the output you can expect from each model. You can extend the output of most models with add-on features.

Read OCR

The Read API analyzes and extracts lines, words, their locations, detected languages, and handwritten style if detected.

Sample document processed using the Document Intelligence Studio:

Layout analysis

The Layout analysis model analyzes and extracts text, tables, selection marks, and other structure elements like titles, section headings, page headers, page footers, and more.

Sample document processed using the Document Intelligence Studio:

Health insurance card

The health insurance card model combines powerful Optical Character Recognition (OCR) capabilities with deep learning models to analyze and extract key information from US health insurance cards.

Sample US health insurance card processed using Document Intelligence Studio:

US tax documents

The US tax document models analyze and extract key fields and line items from a select group of tax documents. The API supports the analysis of English-language US tax documents of various formats and quality including phone-captured images, scanned documents, and digital PDFs. The following models are currently supported:

Model	Description	ModelID
US Tax W-2	Extract taxable compensation details.	prebuilt-tax.us.W-2
US Tax 1040	Extract mortgage interest details.	prebuilt-tax.us.1040(variations)
US Tax 1098	Extract mortgage interest details.	prebuilt-tax.us.1098(variations)
US Tax 1099	Extract income received from sources other than employer.	prebuilt-tax.us.1099(variations)

Sample W-2 document processed using Document Intelligence Studio:

US mortgage documents

The US mortgage document models analyze and extract key fields including borrower, loan, and property information from a select group of mortgage documents. The API supports the analysis of English-language US mortgage documents of various formats and quality including phone-captured images, scanned documents, and digital PDFs. The following models are currently supported:

Model	Description	ModelID
1003 End-User License Agreement (EULA)	Extract loan, borrower, property details.	prebuilt-mortgage.us.1003
1008 Summary document	Extract borrower, seller, property, mortgage, and underwriting details.	prebuilt-mortgage.us.1008
Closing disclosure	Extract closing, transaction costs, and loan details.	prebuilt-mortgage.us.closingDisclosure
Marriage certificate	Extract marriage information details for joint loan applicants.	prebuilt-marriageCertificate
US Tax W-2	Extract taxable compensation details for income verification.	prebuilt-tax.us.W-2

Sample Closing disclosure document processed using Document Intelligence Studio:

Contract

The contract model analyzes and extracts key fields and line items from contractual agreements including parties, jurisdictions, contract ID, and title. The model currently supports English-language contract documents.

Sample contract processed using Document Intelligence Studio:

Invoice

The invoice model automates processing of invoices to extracts customer name, billing address, due date, and amount due, line items, and other key data. Currently, the model supports English, Spanish, German, French, Italian, Portuguese, and Dutch invoices.

Sample invoice processed using Document Intelligence Studio:

Receipt

Use the receipt model to scan sales receipts for merchant name, dates, line items, quantities, and totals from printed and handwritten receipts. The version v3.0 also supports single-page hotel receipt processing.

Sample receipt processed using Document Intelligence Studio:

Identity document (ID)

Use the Identity document (ID) model to process U.S. Driver's Licenses (all 50 states and District of Columbia) and biographical pages from international passports (excluding visa and other travel documents) to extract key fields.

Sample U.S. Driver's License processed using Document Intelligence Studio:

Marriage certificate

Use the marriage certificate model to process U.S. marriage certificates to extract key fields including the individuals, date, and location.

Sample U.S. marriage certificate processed using Document Intelligence Studio:

Credit card

Use the credit card model to process credit and debit cards to extract key fields.

Sample credit card processed using Document Intelligence Studio:

Custom models

Custom models can be broadly classified into two types. Custom classification models that support classification of a "document type" and custom extraction models that can extract a defined schema from a specific document type.

Custom document models analyze and extract data from forms and documents specific to your business. They recognize form fields within your distinct content and extract key-value pairs and table data. You only need one example of the form type to get started.

Version v3.0 and later custom models support signature detection in custom template (form) and cross-page tables in both template and neural models. Signature detection looks for the presence of a signature, not the identity of the person who signs the document. If the model returns unsigned for signature detection, the model didn't find a signature in the defined field.

Sample custom template processed using Document Intelligence Studio:

Custom extraction

Custom extraction model can be custom template. To create a custom extraction model, label a dataset of documents with the values you want extracted and train the model on the labeled dataset. You only need five examples of the same form or document type to get started.

Sample custom extraction processed using Document Intelligence Studio:

Custom classifier

The custom classification model enables you to identify the document type before invoking the extraction model. The classification model is available starting with the 2023-07-31 (GA) API. Training a custom classification model requires at least two distinct classes and a minimum of five samples per class.

Composed models

A composed model is created by taking a collection of custom models and assigning them to a single model built from your form types. You can assign multiple custom models to a composed model called with a single model ID. You can assign up to 200 trained custom models to a single composed model.

Composed model dialog window in Document Intelligence Studio:

Input requirements

• Supported file formats:

  Model	PDF	Image: JPEG/JPG, PNG, BMP, TIFF, HEIF	Microsoft Office: Word (DOCX), Excel (XLSX), PowerPoint (PPTX), HTML
  Read	✔	✔	✔
  Layout	✔	✔	✔ (2024-07-31-preview, 2024-02-29-preview, 2023-10-31-preview)
  General Document	✔	✔
  Prebuilt	✔	✔
  Custom extraction	✔	✔
  Custom classification	✔	✔	✔ (2024-07-31-preview, 2024-02-29-preview)

• For best results, provide one clear photo or high-quality scan per document.

• For PDF and TIFF, up to 2,000 pages can be processed (with a free tier subscription, only the first two pages are processed).

• The file size for analyzing documents is 500 MB for paid (S0) tier and 4 MB for free (F0) tier.

• Image dimensions must be between 50 pixels x 50 pixels and 10,000 pixels x 10,000 pixels.

• If your PDFs are password-locked, you must remove the lock before submission.

• The minimum height of the text to be extracted is 12 pixels for a 1024 x 768 pixel image. This dimension corresponds to about 8 point text at 150 dots per inch (DPI).

• For custom model training, the maximum number of pages for training data is 500 for the custom template model and 50,000 for the custom neural model.

  • For custom extraction model training, the total size of training data is 50 MB for template model and 1 GB for the neural model.

  • For custom classification model training, the total size of training data is 1 GB with a maximum of 10,000 pages. For 2024-07-31-preview and later, the total size of training data is 2 GB with a maximum of 10,000 pages.

Version migration

Learn how to use Document Intelligence v3.0 in your applications by following our Document Intelligence v3.1 migration guide

Model	Description
Document analysis
Layout	Extract text and layout information from documents.
Prebuilt
Invoice	Extract key information from English and Spanish invoices.
Receipt	Extract key information from English receipts.
ID document	Extract key information from US driver licenses and international passports.
Business card	Extract key information from English business cards.
Custom
Custom	Extract data from forms and documents specific to your business. Custom models are trained for your distinct data and use cases.
Composed	Compose a collection of custom models and assign them to a single model built from your form types.

Layout

The Layout API analyzes and extracts text, tables and headers, selection marks, and structure information from documents.

Sample document processed using the Sample Labeling tool:

Invoice

The invoice model analyzes and extracts key information from sales invoices. The API analyzes invoices in various formats and extracts key information such as customer name, billing address, due date, and amount due.

Sample invoice processed using the Sample Labeling tool:

Receipt

• The receipt model analyzes and extracts key information from printed and handwritten sales receipts.

Sample receipt processed using Sample Labeling tool:

ID document

The ID document model analyzes and extracts key information from the following documents:

• U.S. Driver's Licenses (all 50 states and District of Columbia)

• Biographical pages from international passports (excluding visa and other travel documents). The API analyzes identity documents and extracts

Sample U.S. Driver's License processed using the Sample Labeling tool:

Business card

The business card model analyzes and extracts key information from business card images.

Sample business card processed using the Sample Labeling tool:

Custom

• Custom models analyze and extract data from forms and documents specific to your business. The API is a machine-learning program trained to recognize form fields within your distinct content and extract key-value pairs and table data. You only need five examples of the same form type to get started and your custom model can be trained with or without labeled datasets.

Sample custom model processing using the Sample Labeling tool:

Composed custom model

A composed model is created by taking a collection of custom models and assigning them to a single model built from your form types. You can assign multiple custom models to a composed model called with a single model ID. You can assign up to 100 trained custom models to a single composed model.

Composed model dialog window using the Sample Labeling tool:

Model data extraction

Model	Text extraction	Language detection	Selection Marks	Tables	Paragraphs	Paragraph roles	Key-Value pairs	Fields
Layout	✓		✓	✓	✓	✓
Invoice	✓		✓	✓	✓		✓	✓
Receipt	✓				✓			✓
ID Document	✓				✓			✓
Business Card	✓				✓			✓
Custom Form	✓		✓	✓	✓			✓

Input requirements

• Supported file formats:

  Model	PDF	Image: JPEG/JPG, PNG, BMP, TIFF, HEIF	Microsoft Office: Word (DOCX), Excel (XLSX), PowerPoint (PPTX), HTML
  Read	✔	✔	✔
  Layout	✔	✔	✔ (2024-07-31-preview, 2024-02-29-preview, 2023-10-31-preview)
  General Document	✔	✔
  Prebuilt	✔	✔
  Custom extraction	✔	✔
  Custom classification	✔	✔	✔ (2024-07-31-preview, 2024-02-29-preview)

• For best results, provide one clear photo or high-quality scan per document.

• For PDF and TIFF, up to 2,000 pages can be processed (with a free tier subscription, only the first two pages are processed).

• The file size for analyzing documents is 500 MB for paid (S0) tier and 4 MB for free (F0) tier.

• Image dimensions must be between 50 pixels x 50 pixels and 10,000 pixels x 10,000 pixels.

• If your PDFs are password-locked, you must remove the lock before submission.

• The minimum height of the text to be extracted is 12 pixels for a 1024 x 768 pixel image. This dimension corresponds to about 8 point text at 150 dots per inch (DPI).

• For custom model training, the maximum number of pages for training data is 500 for the custom template model and 50,000 for the custom neural model.

  • For custom extraction model training, the total size of training data is 50 MB for template model and 1 GB for the neural model.

  • For custom classification model training, the total size of training data is 1 GB with a maximum of 10,000 pages. For 2024-07-31-preview and later, the total size of training data is 2 GB with a maximum of 10,000 pages.

Version migration

You can learn how to use Document Intelligence v3.0 in your applications by following our Document Intelligence v3.1 migration guide

• Try processing your own forms and documents with the Document Intelligence Studio.

• Complete a Document Intelligence quickstart and get started creating a document processing app in the development language of your choice.

• Try processing your own forms and documents with the Document Intelligence Sample Labeling tool.

• Complete a Document Intelligence quickstart and get started creating a document processing app in the development language of your choice.