Azure Document Intelligence (previously known as Form Recognizer) is a cloud service that uses machine learning to analyze text and structured data from your documents. It includes the following main features:
- Layout - Extract content and structure (ex. words, selection marks, tables) from documents.
- Document - Analyze key-value pairs in addition to general layout from documents.
- Read - Read page information from documents.
- Prebuilt - Extract common field values from select document types (ex. receipts, invoices, business cards, ID documents, U.S. W-2 tax documents, among others) using prebuilt models.
- Custom - Build custom models from your own data to extract tailored field values in addition to general layout from documents.
- Classifiers - Build custom classification models that combine layout and language features to accurately detect and identify documents you process within your application.
- Add-on capabilities - Extract barcodes/QR codes, formulas, font/style, etc. or enable high resolution mode for large documents with optional parameters.
Source code | Package (PyPI) | Package (Conda) | API reference documentation | Product documentation | Samples
This package supports the following service API versions: 2.0, 2.1, 2022-08-31 and 2023-07-31. Service API version 2023-10-31-preview and later are supported in package azure-ai-documentintelligence
. Please refer this doc for migration details.
- Python 3.8 or later is required to use this package.
- You must have an Azure subscription and a Cognitive Services or Form Recognizer resource to use this package.
Install the Azure Form Recognizer client library for Python with pip:
pip install azure-ai-formrecognizer
Note: This version of the client library defaults to the
2023-07-31
version of the service.
This table shows the relationship between SDK versions and supported API versions of the service:
SDK version | Supported API version of service |
---|---|
3.3.X - Latest GA release | 2.0, 2.1, 2022-08-31, 2023-07-31 (default) |
3.2.X | 2.0, 2.1, 2022-08-31 (default) |
3.1.X | 2.0, 2.1 (default) |
3.0.0 | 2.0 |
Note: Starting with version
3.2.X
, a new set of clients were introduced to leverage the newest features of the Document Intelligence service. Please see the Migration Guide for detailed instructions on how to update application code from client library version3.1.X
or lower to the latest version. Additionally, see the Changelog for more detailed information. The below table describes the relationship of each client and its supported API version(s):
API version | Supported clients |
---|---|
2023-07-31 | DocumentAnalysisClient and DocumentModelAdministrationClient |
2022-08-31 | DocumentAnalysisClient and DocumentModelAdministrationClient |
2.1 | FormRecognizerClient and FormTrainingClient |
2.0 | FormRecognizerClient and FormTrainingClient |
Document Intelligence supports both multi-service and single-service access. Create a Cognitive Services resource if you plan to access multiple cognitive services under a single endpoint/key. For Document Intelligence access only, create a Form Recognizer resource. Please note that you will need a single-service resource if you intend to use Azure Active Directory authentication.
You can create either resource using:
- Option 1: Azure Portal.
- Option 2: Azure CLI.
Below is an example of how you can create a Form Recognizer resource using the CLI:
# Create a new resource group to hold the Form Recognizer resource
# if using an existing resource group, skip this step
az group create --name <your-resource-name> --location <location>
# Create form recognizer
az cognitiveservices account create \
--name <your-resource-name> \
--resource-group <your-resource-group-name> \
--kind FormRecognizer \
--sku <sku> \
--location <location> \
--yes
For more information about creating the resource or how to get the location and sku information see here.
In order to interact with the Document Intelligence service, you will need to create an instance of a client. An endpoint and credential are necessary to instantiate the client object.
You can find the endpoint for your Form Recognizer resource using the Azure Portal or Azure CLI:
# Get the endpoint for the Form Recognizer resource
az cognitiveservices account show --name "resource-name" --resource-group "resource-group-name" --query "properties.endpoint"
Either a regional endpoint or a custom subdomain can be used for authentication. They are formatted as follows:
Regional endpoint: https://<region>.api.cognitive.microsoft.com/
Custom subdomain: https://<resource-name>.cognitiveservices.azure.com/
A regional endpoint is the same for every resource in a region. A complete list of supported regional endpoints can be consulted here. Please note that regional endpoints do not support AAD authentication.
A custom subdomain, on the other hand, is a name that is unique to the Form Recognizer resource. They can only be used by single-service resources.
The API key can be found in the Azure Portal or by running the following Azure CLI command:
az cognitiveservices account keys list --name "<resource-name>" --resource-group "<resource-group-name>"
To use an API key as the credential
parameter,
pass the key as a string into an instance of AzureKeyCredential.
from azure.core.credentials import AzureKeyCredential
from azure.ai.formrecognizer import DocumentAnalysisClient
endpoint = "https://<my-custom-subdomain>.cognitiveservices.azure.com/"
credential = AzureKeyCredential("<api_key>")
document_analysis_client = DocumentAnalysisClient(endpoint, credential)
AzureKeyCredential
authentication is used in the examples in this getting started guide, but you can also
authenticate with Azure Active Directory using the azure-identity library.
Note that regional endpoints do not support AAD authentication. Create a custom subdomain
name for your resource in order to use this type of authentication.
To use the DefaultAzureCredential type shown below, or other credential types provided
with the Azure SDK, please install the azure-identity
package:
pip install azure-identity
You will also need to register a new AAD application and grant access to Document Intelligence by assigning the "Cognitive Services User"
role to your service principal.
Once completed, set the values of the client ID, tenant ID, and client secret of the AAD application as environment variables:
AZURE_CLIENT_ID
, AZURE_TENANT_ID
, AZURE_CLIENT_SECRET
.
"""DefaultAzureCredential will use the values from these environment
variables: AZURE_CLIENT_ID, AZURE_TENANT_ID, AZURE_CLIENT_SECRET
"""
from azure.ai.formrecognizer import DocumentAnalysisClient
from azure.identity import DefaultAzureCredential
endpoint = os.environ["AZURE_FORM_RECOGNIZER_ENDPOINT"]
credential = DefaultAzureCredential()
document_analysis_client = DocumentAnalysisClient(endpoint, credential)
DocumentAnalysisClient
provides operations for analyzing input documents using prebuilt and custom models through the begin_analyze_document
and begin_analyze_document_from_url
APIs.
Use the model_id
parameter to select the type of model for analysis. See a full list of supported models here.
The DocumentAnalysisClient
also provides operations for classifying documents through the begin_classify_document
and begin_classify_document_from_url
APIs.
Custom classification models can classify each page in an input file to identify the document(s) within and can also identify multiple documents or multiple instances of a single document within an input file.
Sample code snippets are provided to illustrate using a DocumentAnalysisClient here. More information about analyzing documents, including supported features, locales, and document types can be found in the service documentation.
DocumentModelAdministrationClient
provides operations for:
- Building custom models to analyze specific fields you specify by labeling your custom documents. A
DocumentModelDetails
is returned indicating the document type(s) the model can analyze, as well as the estimated confidence for each field. See the service documentation for a more detailed explanation. - Creating a composed model from a collection of existing models.
- Managing models created in your account.
- Listing operations or getting a specific model operation created within the last 24 hours.
- Copying a custom model from one Form Recognizer resource to another.
- Build and manage a custom classification model to classify the documents you process within your application.
Please note that models can also be built using a graphical user interface such as Document Intelligence Studio.
Sample code snippets are provided to illustrate using a DocumentModelAdministrationClient here.
Long-running operations are operations which consist of an initial request sent to the service to start an operation, followed by polling the service at intervals to determine whether the operation has completed or failed, and if it has succeeded, to get the result.
Methods that analyze documents, build models, or copy/compose models are modeled as long-running operations.
The client exposes a begin_<method-name>
method that returns an LROPoller
or AsyncLROPoller
. Callers should wait
for the operation to complete by calling result()
on the poller object returned from the begin_<method-name>
method.
Sample code snippets are provided to illustrate using long-running operations below.
The following section provides several code snippets covering some of the most common Document Intelligence tasks, including:
- Extract Layout
- Using the General Document Model
- Using Prebuilt Models
- Build a Custom Model
- Analyze Documents Using a Custom Model
- Manage Your Models
- Classify Documents
- Add-on capabilities
Extract text, selection marks, text styles, and table structures, along with their bounding region coordinates, from documents.
from azure.core.credentials import AzureKeyCredential
from azure.ai.formrecognizer import DocumentAnalysisClient
endpoint = os.environ["AZURE_FORM_RECOGNIZER_ENDPOINT"]
key = os.environ["AZURE_FORM_RECOGNIZER_KEY"]
document_analysis_client = DocumentAnalysisClient(
endpoint=endpoint, credential=AzureKeyCredential(key)
)
with open(path_to_sample_documents, "rb") as f:
poller = document_analysis_client.begin_analyze_document(
"prebuilt-layout", document=f
)
result = poller.result()
for idx, style in enumerate(result.styles):
print(
"Document contains {} content".format(
"handwritten" if style.is_handwritten else "no handwritten"
)
)
for page in result.pages:
print("----Analyzing layout from page #{}----".format(page.page_number))
print(
"Page has width: {} and height: {}, measured with unit: {}".format(
page.width, page.height, page.unit
)
)
for line_idx, line in enumerate(page.lines):
words = line.get_words()
print(
"...Line # {} has word count {} and text '{}' within bounding polygon '{}'".format(
line_idx,
len(words),
line.content,
line.polygon,
)
)
for word in words:
print(
"......Word '{}' has a confidence of {}".format(
word.content, word.confidence
)
)
for selection_mark in page.selection_marks:
print(
"...Selection mark is '{}' within bounding polygon '{}' and has a confidence of {}".format(
selection_mark.state,
selection_mark.polygon,
selection_mark.confidence,
)
)
for table_idx, table in enumerate(result.tables):
print(
"Table # {} has {} rows and {} columns".format(
table_idx, table.row_count, table.column_count
)
)
for region in table.bounding_regions:
print(
"Table # {} location on page: {} is {}".format(
table_idx,
region.page_number,
region.polygon,
)
)
for cell in table.cells:
print(
"...Cell[{}][{}] has content '{}'".format(
cell.row_index,
cell.column_index,
cell.content,
)
)
for region in cell.bounding_regions:
print(
"...content on page {} is within bounding polygon '{}'".format(
region.page_number,
region.polygon,
)
)
print("----------------------------------------")
Analyze key-value pairs, tables, styles, and selection marks from documents using the general document model provided by the Document Intelligence service.
Select the General Document Model by passing model_id="prebuilt-document"
into the begin_analyze_document
method:
from azure.core.credentials import AzureKeyCredential
from azure.ai.formrecognizer import DocumentAnalysisClient
endpoint = os.environ["AZURE_FORM_RECOGNIZER_ENDPOINT"]
key = os.environ["AZURE_FORM_RECOGNIZER_KEY"]
document_analysis_client = DocumentAnalysisClient(
endpoint=endpoint, credential=AzureKeyCredential(key)
)
with open(path_to_sample_documents, "rb") as f:
poller = document_analysis_client.begin_analyze_document(
"prebuilt-document", document=f
)
result = poller.result()
for style in result.styles:
if style.is_handwritten:
print("Document contains handwritten content: ")
print(",".join([result.content[span.offset:span.offset + span.length] for span in style.spans]))
print("----Key-value pairs found in document----")
for kv_pair in result.key_value_pairs:
if kv_pair.key:
print(
"Key '{}' found within '{}' bounding regions".format(
kv_pair.key.content,
kv_pair.key.bounding_regions,
)
)
if kv_pair.value:
print(
"Value '{}' found within '{}' bounding regions\n".format(
kv_pair.value.content,
kv_pair.value.bounding_regions,
)
)
for page in result.pages:
print("----Analyzing document from page #{}----".format(page.page_number))
print(
"Page has width: {} and height: {}, measured with unit: {}".format(
page.width, page.height, page.unit
)
)
for line_idx, line in enumerate(page.lines):
words = line.get_words()
print(
"...Line # {} has {} words and text '{}' within bounding polygon '{}'".format(
line_idx,
len(words),
line.content,
line.polygon,
)
)
for word in words:
print(
"......Word '{}' has a confidence of {}".format(
word.content, word.confidence
)
)
for selection_mark in page.selection_marks:
print(
"...Selection mark is '{}' within bounding polygon '{}' and has a confidence of {}".format(
selection_mark.state,
selection_mark.polygon,
selection_mark.confidence,
)
)
for table_idx, table in enumerate(result.tables):
print(
"Table # {} has {} rows and {} columns".format(
table_idx, table.row_count, table.column_count
)
)
for region in table.bounding_regions:
print(
"Table # {} location on page: {} is {}".format(
table_idx,
region.page_number,
region.polygon,
)
)
for cell in table.cells:
print(
"...Cell[{}][{}] has content '{}'".format(
cell.row_index,
cell.column_index,
cell.content,
)
)
for region in cell.bounding_regions:
print(
"...content on page {} is within bounding polygon '{}'\n".format(
region.page_number,
region.polygon,
)
)
print("----------------------------------------")
- Read more about the features provided by the
prebuilt-document
model here.
Extract fields from select document types such as receipts, invoices, business cards, identity documents, and U.S. W-2 tax documents using prebuilt models provided by the Document Intelligence service.
For example, to analyze fields from a sales receipt, use the prebuilt receipt model provided by passing model_id="prebuilt-receipt"
into the begin_analyze_document
method:
from azure.core.credentials import AzureKeyCredential
from azure.ai.formrecognizer import DocumentAnalysisClient
endpoint = os.environ["AZURE_FORM_RECOGNIZER_ENDPOINT"]
key = os.environ["AZURE_FORM_RECOGNIZER_KEY"]
document_analysis_client = DocumentAnalysisClient(
endpoint=endpoint, credential=AzureKeyCredential(key)
)
with open(path_to_sample_documents, "rb") as f:
poller = document_analysis_client.begin_analyze_document(
"prebuilt-receipt", document=f, locale="en-US"
)
receipts = poller.result()
for idx, receipt in enumerate(receipts.documents):
print(f"--------Analysis of receipt #{idx + 1}--------")
print(f"Receipt type: {receipt.doc_type if receipt.doc_type else 'N/A'}")
merchant_name = receipt.fields.get("MerchantName")
if merchant_name:
print(
f"Merchant Name: {merchant_name.value} has confidence: "
f"{merchant_name.confidence}"
)
transaction_date = receipt.fields.get("TransactionDate")
if transaction_date:
print(
f"Transaction Date: {transaction_date.value} has confidence: "
f"{transaction_date.confidence}"
)
if receipt.fields.get("Items"):
print("Receipt items:")
for idx, item in enumerate(receipt.fields.get("Items").value):
print(f"...Item #{idx + 1}")
item_description = item.value.get("Description")
if item_description:
print(
f"......Item Description: {item_description.value} has confidence: "
f"{item_description.confidence}"
)
item_quantity = item.value.get("Quantity")
if item_quantity:
print(
f"......Item Quantity: {item_quantity.value} has confidence: "
f"{item_quantity.confidence}"
)
item_price = item.value.get("Price")
if item_price:
print(
f"......Individual Item Price: {item_price.value} has confidence: "
f"{item_price.confidence}"
)
item_total_price = item.value.get("TotalPrice")
if item_total_price:
print(
f"......Total Item Price: {item_total_price.value} has confidence: "
f"{item_total_price.confidence}"
)
subtotal = receipt.fields.get("Subtotal")
if subtotal:
print(f"Subtotal: {subtotal.value} has confidence: {subtotal.confidence}")
tax = receipt.fields.get("TotalTax")
if tax:
print(f"Total tax: {tax.value} has confidence: {tax.confidence}")
tip = receipt.fields.get("Tip")
if tip:
print(f"Tip: {tip.value} has confidence: {tip.confidence}")
total = receipt.fields.get("Total")
if total:
print(f"Total: {total.value} has confidence: {total.confidence}")
print("--------------------------------------")
You are not limited to receipts! There are a few prebuilt models to choose from, each of which has its own set of supported fields. See other supported prebuilt models here.
Build a custom model on your own document type. The resulting model can be used to analyze values from the types of documents it was trained on. Provide a container SAS URL to your Azure Storage Blob container where you're storing the training documents.
More details on setting up a container and required file structure can be found in the service documentation.
from azure.ai.formrecognizer import (
DocumentModelAdministrationClient,
ModelBuildMode,
)
from azure.core.credentials import AzureKeyCredential
endpoint = os.environ["AZURE_FORM_RECOGNIZER_ENDPOINT"]
key = os.environ["AZURE_FORM_RECOGNIZER_KEY"]
container_sas_url = os.environ["CONTAINER_SAS_URL"]
document_model_admin_client = DocumentModelAdministrationClient(
endpoint, AzureKeyCredential(key)
)
poller = document_model_admin_client.begin_build_document_model(
ModelBuildMode.TEMPLATE,
blob_container_url=container_sas_url,
description="my model description",
)
model = poller.result()
print(f"Model ID: {model.model_id}")
print(f"Description: {model.description}")
print(f"Model created on: {model.created_on}")
print(f"Model expires on: {model.expires_on}")
print("Doc types the model can recognize:")
for name, doc_type in model.doc_types.items():
print(
f"Doc Type: '{name}' built with '{doc_type.build_mode}' mode which has the following fields:"
)
for field_name, field in doc_type.field_schema.items():
print(
f"Field: '{field_name}' has type '{field['type']}' and confidence score "
f"{doc_type.field_confidence[field_name]}"
)
Analyze document fields, tables, selection marks, and more. These models are trained with your own data, so they're tailored to your documents. For best results, you should only analyze documents of the same document type that the custom model was built with.
from azure.core.credentials import AzureKeyCredential
from azure.ai.formrecognizer import DocumentAnalysisClient
endpoint = os.environ["AZURE_FORM_RECOGNIZER_ENDPOINT"]
key = os.environ["AZURE_FORM_RECOGNIZER_KEY"]
model_id = os.getenv("CUSTOM_BUILT_MODEL_ID", custom_model_id)
document_analysis_client = DocumentAnalysisClient(
endpoint=endpoint, credential=AzureKeyCredential(key)
)
# Make sure your document's type is included in the list of document types the custom model can analyze
with open(path_to_sample_documents, "rb") as f:
poller = document_analysis_client.begin_analyze_document(
model_id=model_id, document=f
)
result = poller.result()
for idx, document in enumerate(result.documents):
print(f"--------Analyzing document #{idx + 1}--------")
print(f"Document has type {document.doc_type}")
print(f"Document has document type confidence {document.confidence}")
print(f"Document was analyzed with model with ID {result.model_id}")
for name, field in document.fields.items():
field_value = field.value if field.value else field.content
print(
f"......found field of type '{field.value_type}' with value '{field_value}' and with confidence {field.confidence}"
)
# iterate over tables, lines, and selection marks on each page
for page in result.pages:
print(f"\nLines found on page {page.page_number}")
for line in page.lines:
print(f"...Line '{line.content}'")
for word in page.words:
print(f"...Word '{word.content}' has a confidence of {word.confidence}")
if page.selection_marks:
print(f"\nSelection marks found on page {page.page_number}")
for selection_mark in page.selection_marks:
print(
f"...Selection mark is '{selection_mark.state}' and has a confidence of {selection_mark.confidence}"
)
for i, table in enumerate(result.tables):
print(f"\nTable {i + 1} can be found on page:")
for region in table.bounding_regions:
print(f"...{region.page_number}")
for cell in table.cells:
print(
f"...Cell[{cell.row_index}][{cell.column_index}] has text '{cell.content}'"
)
print("-----------------------------------")
Alternatively, a document URL can also be used to analyze documents using the begin_analyze_document_from_url
method.
document_url = "<url_of_the_document>"
poller = document_analysis_client.begin_analyze_document_from_url(model_id=model_id, document_url=document_url)
result = poller.result()
Manage the custom models attached to your account.
from azure.ai.formrecognizer import DocumentModelAdministrationClient
from azure.core.credentials import AzureKeyCredential
from azure.core.exceptions import ResourceNotFoundError
endpoint = "https://<my-custom-subdomain>.cognitiveservices.azure.com/"
credential = AzureKeyCredential("<api_key>")
document_model_admin_client = DocumentModelAdministrationClient(endpoint, credential)
account_details = document_model_admin_client.get_resource_details()
print("Our account has {} custom models, and we can have at most {} custom models".format(
account_details.custom_document_models.count, account_details.custom_document_models.limit
))
# Here we get a paged list of all of our models
models = document_model_admin_client.list_document_models()
print("We have models with the following ids: {}".format(
", ".join([m.model_id for m in models])
))
# Replace with the custom model ID from the "Build a model" sample
model_id = "<model_id from the Build a Model sample>"
custom_model = document_model_admin_client.get_document_model(model_id=model_id)
print("Model ID: {}".format(custom_model.model_id))
print("Description: {}".format(custom_model.description))
print("Model created on: {}\n".format(custom_model.created_on))
# Finally, we will delete this model by ID
document_model_admin_client.delete_document_model(model_id=custom_model.model_id)
try:
document_model_admin_client.get_document_model(model_id=custom_model.model_id)
except ResourceNotFoundError:
print("Successfully deleted model with id {}".format(custom_model.model_id))
Document Intelligence supports more sophisticated analysis capabilities. These optional features can be enabled and disabled depending on the scenario of the document extraction.
The following add-on capabilities are available for 2023-07-31 (GA) and later releases:
Note that some add-on capabilities will incur additional charges. See pricing: https://azure.microsoft.com/pricing/details/ai-document-intelligence/.
Form Recognizer client library will raise exceptions defined in Azure Core. Error codes and messages raised by the Document Intelligence service can be found in the service documentation.
This library uses the standard logging library for logging.
Basic information about HTTP sessions (URLs, headers, etc.) is logged at INFO
level.
Detailed DEBUG
level logging, including request/response bodies and unredacted
headers, can be enabled on the client or per-operation with the logging_enable
keyword argument.
See full SDK logging documentation with examples here.
Optional keyword arguments can be passed in at the client and per-operation level. The azure-core reference documentation describes available configurations for retries, logging, transport protocols, and more.
See the Sample README for several code snippets illustrating common patterns used in the Form Recognizer Python API.
For more extensive documentation on Azure AI Document Intelligence, see the Document Intelligence documentation on docs.microsoft.com.
This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit cla.microsoft.com.
When you submit a pull request, a CLA-bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., label, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.
This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.