Document AI release notes

This page documents production updates to Document AI. We recommend that Document AI developers periodically check this list for any new announcements.

You can see the latest product updates for all of Google Cloud on the Google Cloud page, browse and filter all release notes in the Google Cloud console, or programmatically access release notes in BigQuery.

To get the latest product updates delivered to you, add the URL of this page to your feed reader, or add the feed URL directly: https://cloud.google.com/feeds/duai-release-notes.xml

February 29, 2024

v1

The Custom Extractor supports three levels of nesting so you can easily extract structured data from complex documents and tables (earnings reports, tax forms, invoices, resumes, etc.). Learn how to use three levels of nesting.

The Custom Extractor with generative AI is now available in the asia-southeast1 (Singapore) regions. For more information, see Custom processors.

See the model type, generative or custom, powering a Custom Extractor processor version by getting the model type from the processorVersions API.

February 16, 2024

v1

Enterprise Document OCR version 2.0, pretrained-ocr-v2.0-2023-06-02, is now Generally Available and ready for production workloads.

Please migrate OCR workloads to this new processor version.

January 09, 2024

v1

The Custom Extractor with generative AI has General Availability and is ready for production workloads. For more information, see the Custom Extractor with generative AI or check out the demo.

To better support production workloads, we reduced prices for the Custom Extractor, Custom Classifier, Custom Splitter, and Form Parser. For more information, see Document AI pricing.

Developers can now specify pages Document AI should process within a document. For more information, see IndividualPageSelector within V1 API ProcessOptions.

December 20, 2023

v1beta3

Custom Extractor supports fine tuning (Preview) so that you can customize foundation model results for user specific documents. This feature is available in the US region. For more information, see Fine tune and train by document type.

Custom Extractor with genAI is now available in the EU and northamerica-northeast1 regions. For more information, see Custom processors.

You can now demo genAI-powered extraction results within Custom Extractor along with output from other Document AI products such as OCR, Form Parser, and ID processing.

December 07, 2023

v1

Enterprise Document OCR version 2.0, pretrained-ocr-v2.0-2023-06-02, has an upgraded OCR engine and model improvements. This upgrade better supports high-volume workloads.

September 25, 2023

v1

We are launching an RC version of the pretrained-invoice-v1.5-2023-09-15 invoice processor. It includes:

  • Improved base-entity extraction model for documents in English.
  • Line-item grouping quality improvements.
  • Better support for multi-line, multi-segment entities such as addresses and line-item descriptions.
  • Enforcement of occurrence type OPTIONAL_ONCE / REQUIRED_ONCE for properties of nested entities.
  • Updated OCR engine.

September 21, 2023

v1

Launched Document AI Enterprise Document OCR v2.0 and OCR add ons in Preview.

Enterprise Document OCR launched a Release Candidate, pretrained-ocr-v2.0-2023-06-02, which includes:

  • Upgraded OCR model, optimized for various document use cases.
  • Visual-element detector for boxed characters, which can increase quality up to 10% for documents with text boxes.

For more details, see the documentation, including the user guide.

OCR add ons are available from the Enterprise Document OCR processor when using pretrained-ocr-v2.0-2023-06-02. These include:

  • Checkbox extraction: Detects and extracts status (marked/unmarked) in the Enterprise Document OCR response.
  • Math OCR: Identifies, recognizes, and extracts formulas from documents in LaTeX output format.
  • Font-style detection: Identifies word-level font properties, including type, style, handwriting, weight, and color.

For more details, see the documentation.

August 25, 2023

v1

Document AI Workbench is now powered by generative AI with two feature launches:

Document AI Workbench Summarizer is in Preview:

  • The Summarizer provides summaries for documents up to 250 pages long.
  • You can customize summaries based on your preferences for length (brief, moderate, comprehensive) and format (paragraph, bullet points).
  • See the user guide for more information.

Document AI Workbench custom extractor is in preview:

  • Custom extractor with generative AI can help extract data from documents with free-form text (e.g., contracts) and complex layouts (e.g., invoices, W2s, bills of lading).
  • The pretrained processor version, which uses generative AI, can be used out of the box without any training. Post a document to the endpoint with a list of fields to get structured data.
  • Customize results by confirming content in about five documents. Workbench leverages the examples to improve accuracy using few-shot prediction.
  • Extract information from documents up to 200 pages long through the asynchronous API.
  • To get started, create or use an existing custom extractor to leverage a processor version.
  • See the how-to guide, labeling best practices, and training use cases.
  • Current limitations of generative AI extraction within the custom extractor:
    • Only the English language is supported.
    • Region availability is currently only in the US.
    • While in preview, we recommend that you only extract up to 50 entities per endpoint with generative AI.
    • When uploading a sample document to define fields and preview results on the Get started page, there can be long latencies. We're working to reduce this latency.

In addition, template-based training is available in GA within the custom extractor:

  • Template-based training provides accurate predictions for documents with no layout variation (such as an application form).
  • Only six labeled documents are needed to train and use a template-based processor version.
  • See the user guide and training use cases.

August 18, 2023

v1

Expense processor

A new RC version pretrained-expense-v1.3.1-2023-08-11 of the Expense processor is now available in the asia-southeast1 region for Expense Parser customers.

This release includes an improved region-based normalization, which results in an average improvement of up to a 15% accuracy on normalized date and currency entities over the current stable version.

August 01, 2023

v1

Launched the following Document AI Workbench features:

  • Create and train models programmatically with more public APIs, including:

  • Selective labeling within Custom Document Extractor (CDE) helps you prepare a diverse set of training and test documents. Import 125+ documents to a CDE dataset, then CDE recommends documents you should label based on clustering results. Selective labeling only supports new documents imported into your dataset. If you would like recommendations for documents already in your dataset, then delete and re-import them.

  • Quick Tables within CDE helps you train models faster by labeling a table in bulk by applying the first row pattern to the rest of the table.

  • Get started with a new processor by using a default storage option for your dataset. You can still configure your own Cloud Storage location using advanced options.

July 28, 2023

v1

Launched an update to the RC release pretrained-invoice-v1.4-2022-10-21 of the invoice processor, available to all Document AI users.

This release includes the features of the Stable version, with an average improvement of 0.03 to 0.05 micro F1 (approximately 5% to 9%) on line item entities.

July 18, 2023

v1

The following Form Parser (pretrained-form-parser-v2.0-2022-11-10) features are Generally Available (GA):

  • General field extraction: You can extract 11 different types of entities from documents.
  • Enhanced check-box detection.
  • Internationalization support that covers more than 200 languages.
  • Upgraded key-value pair (KVP) detection model.

Form parser v2.1 (pretrained-form-parser-v2.1-2023-06-26) is in Public Preview, which uses our native PDF text extraction model on PDF documents.

The Form Parser features has the following limitations:

  • Check-box doesn't support radio buttons and might not reliably parse all selection marks or keyless checkboxes.
  • If there is a key without a value, the model might not parse it.
  • The quality of KVP parsing might be higher for Latin languages than others.
  • For tables, only simple tables are supported (no support for merged cells).

July 17, 2023

v1

The Custom Document Splitter (CDS) within Document AI Workbench is now Generally Available (GA) for production use cases to split and classify multiple documents within a single file. With this release, all Workbench processors currently offered (Custom Document Extractor, Custom Document Classifier, and Custom Document Splitter) are available in GA.

Launched the following features for CDS:

  • CDS now supports up to 1,000-page documents for async/batch prediction and up to 200-page documents when importing, labeling, training, or evaluating.
  • CDS model evaluation for document split and classification
  • Prepare a CDS training dataset faster by bulk labeling documents at import across multiple folders.

Released the the following enhancements for CDS:

  • Improved labeling and evaluation experience with the ability to review overall document splits and classifications while viewing individual pages in a side-by-side view.
  • Document names are now used in error messaging to improve troubleshooting.
  • Hyphens are allowed in schema names.

June 29, 2023

v1

Identity Document AI (IDAI) photo copy detection in ID proofing (Preview)

Updated the pretrained-id-proofing-v1.1-2023-05-18 ID proofing processor for all Document AI users.

This processor includes a new output entity fraud_signals_photocopy_detection that signals if an attached image might be a photocopy. The entity can be one of the following values: POSSIBLE_PHOTOCOPY, PASS, or INCONCLUSIVE.

June 28, 2023

v1

The following document OCR features are Generally Available (GA). Use document OCR's configurations to optimize for stability, quality, and specific response requirements.

  • Intelligent document-quality analysis
  • Native text from digital PDF
  • Symbol-level extraction
  • Language hints

Support for DOCX is in Preview. You can synchronously process DOCX files that are up to 15 pages, or asynchronously process DOCX files that are up to 30. For access, send us a request.

Added fixes to our doc.proto-to-vision.proto conversion tool, which facilitates migration from Vision API TextDetection to document OCR.

The document OCR native text from digital PDF feature contains the following known issues:

  • For a small number of documents, word order in lines of text that are reported by native text extraction might be inaccurate.
  • Invisible text that is embedded in a native PDF might be extracted.
  • Japanese documents that contain currency symbols, such as Yen, might be incorrectly extracted as /.
  • Apostrophe symbols might be missing in word and/or line results.
  • Native text extraction might report different word and/or line results compared to image-based OCR on an identical document.

April 25, 2023

v1

Launched the following features to improve the usability of the Document AI Workbench Custom Document Extractor (CDE):

  • CDE now supports an additional 42 global languages.
  • CDE lets you import processor versions across projects and processors to easily manage development and production environments.
  • CDE can automatically label documents in a dataset by using a deployed processor version to help you quickly prepare training data.

Document AI Workbench Custom Document Extractor (CDE) has also made the following enhancements:

  • The asynchronous prediction API can now extract data from documents up to 200 pages long.
  • Improved the accuracy of extracting checkboxes.

April 17, 2023

v1

Identity Document AI (IDAI) pricing change

We are changing the price of our US-related Identity Document processors. The new price is on the pricing page.

March 27, 2023

v1beta3

The Document AI OCR Processor (Doc OCR) now has the following features:

  • The OCR Processor supports language hints. The OCR engine prefers your specified languages over inferred languages. To use this feature, set process_options.ocr_config.hints.language_hints with a list of BCP-47 language codes in your API request to the OCR Processor.
  • The OCR Processor supports the option to populate symbol-level data in the document response. If enabled, the field document.pages.symbols is populated. To use this feature, set process_options.ocr_config.enable_symbol=true in your API request to the OCR Processor.
  • A proto converter tool that converts a Document proto to an AnnotateFileResponse proto. This conversion lets you compare the responses between the Document AI OCR processor with the Vision API, which can help you migrate to the Document AI OCR processor from Vision API with minimal downstream changes. For details, see Document AI Toolbox.
  • The OCR Processor supports a heuristics layout detection algorithm, which serves as an alternative to the current ML-based layout detection algorithm. You can choose the layout algorithm that best suits your needs. To use this feature, set process_options.ocr_config.advanced_ocr_options= legacy_layout in your API request to the OCR Processor.

For the Document AI OCR Processor (Doc OCR), you can enable document quality assessments for all processor versions instead of a specific processor version, such as pretrained-ocr-v1.1-2022-09-12. If you enable document quality assessment, Doc OCR produces a quality score that's based on the document's readability. Quality scores range from 0 to 1, where 1 is perfect quality. Quality scores are returned in the image_quality_scores field on the Page object. All detected issues are labeled as quality or defect and sorted in descending order by confidence value. To use this feature, set process_options.ocr_config.enable_image_quality_scores= true in your API request to the OCR Processor.

February 21, 2023

v1

This launch upgrades the lifecycle stage of the Custom Document Extractor (CDE) component of the DocAI Workbench from Public Preview to Generally Available (GA). CDE covers essential workflows for developing custom document extraction processors with end-to-end UI support:

  • Data import
  • Schema creation and annotation
  • Processor model training
  • Evaluation and troubleshooting
  • Model deployment and version management
  • Human-in-the-loop (HITL) integration for "last-mile" processor quality assurance

Notable new Generally Available Custom Document Extractor (CDE) features include:

  • Public APIs
  • Automatic schema label creation from pre-labeled documents
  • Schema label data type and occurrence editable pre-training
  • New DocAI Toolkit with a labeled document converter

The following features have been upgraded:

  • Processor gallery
  • Schema editor
  • Labeling UI
  • Training pipeline
  • Manage versions table

January 10, 2023

v1beta3 & v1

The Form Parser Release Candidate version has been renamed to pretrained-form-parser-v2.0-2022-11-10. See Document AI release notes--December 12, 2022 for more information about this release.

December 22, 2022

v1beta3

We are launching a public preview version of the purchase order processor, pretrained-purchase-order-v1.1-2022-06-17, with the following features:

  • Support for uptraining to improve, add, and remove entities in the schema.
  • Support for uptraining to add support for unsupported languages.
  • Improvements to overall performance.

December 19, 2022

v1beta3

The Document AI OCR Processor has the following new features:

  • The OCR Processor now supports extracting embedded text from digital PDFs in public preview. A fallback to the optical OCR model is automatically triggered to extract text in the regions when the PDF being processed contains non-digital text. To opt into this feature, set process_options.ocr_config.enable_native_pdf_parsing=true in your API request to the OCR Processor.

  • Added advanced versioning support to the Document AI OCR, which enables OCR users to pin to a historical model version. When enabled, OCR outputs are guaranteed to be consistent and virtually frozen, with zero behavioral drifts. To enable advanced versioning, select the release candidate version pretrained-ocr-v1.2-2022-11-10 in your Document AI console.

Known issues with the digital PDF feature of the Document AI OCR Processor:

  • On a small number of documents, the word ordering within lines of text as reported by native text extraction might be wrong.

  • On certain documents, invisible text embedded in a native PDF may be reported.

  • On certain Japanese documents, currency symbols such as Yen might be incorrectly extracted as /.

  • On certain documents, apostrophe symbols may be missing in word/line results.

  • On certain documents, native text extraction might report different word/line results than those obtained by image-based OCR on an identical document.

December 15, 2022

v1

We are launching the Release candidate version, pretrained-utility-v1.2-2022-12-15 of the utility processor. This version includes the following features:

  • Removal of three entities from schema: delivery_date, receiver_email, receiver_phone
  • Improvements to overall performance

The utility processor version pretrained-utility-v1.1-2021-04-09 continues to serve as the current, latest Stable version.

December 12, 2022

v1beta3

The Form Parser now supports Generic Entity Extraction in Public Preview, covering the following entity types:

  • email: email address
  • phone: phone number
  • url: website URLs
  • date_time: partial or full date/time/period
  • address: full address or street address in a single line
  • person: partial or full name of a person
  • organization: full name of an organization
  • quantity: a number specifying quantity or percentage
  • price: a number specifying monetary amount
  • id: a number specifying identity
  • page_number: a number specifying page number

The Form Parser has the following feature enhancements:

  • The Form Parser key-value pair (entity and checkbox) extraction and table extraction now support 200+ languages that are supported by the underlying multi-language OCR model. This language expansion is in Public Preview, with key-value pair internationalization backed by quality benchmarks in selected languages such as Simplified Chinese, Traditional Chinese, Japanese, and Korean.

  • Table extraction in Form Parser is now powered by an enhanced vision-based table parsing model.

These enhanced features are automatically enabled for Form Parser processor version pretrained-parser-v2.0-2022-11-10 and all future versions. Note that this is a Release Candidate version, which is subject to further changes before graduating to the Stable version.

December 08, 2022

v1

Invoice Parser and Expense Parser are now available in the following single-region locations:

  • australia-southeast1 (Sydney)
  • northamerica-northeast1 (Montréal)

November 21, 2022

v1

Expense Parser Releases

As of November 18, 2022, for the Expense Parser, we have promoted our v1.3 Release Candidate version to a Stable version so that more customers can use it confidently. 

New Stable version

Features in the new Stable Expense Parser, pretrained-expense-v1.3-2022-07-15:

  • Support for a new language, Japanese, which has been requested by multiple customers.

  • Better entity performance

  • Addition of 3 new entity types (line_item/quantity, payment_typecredit_card_last_four_digits)

  • Better support for hotel and car-rental related expenses 

New Release Candidate version

Along with this Stable version, we are also launching a new Release Candidate version of the Expense Parser, pretrained-expense-v1.4-2022-11-18, with the following new features, in addition to the features in the Stable version:

  • Improvements to overall performance

  • Support for two (2) new languages, Italian and Portuguese

  • Support for Uptraining to improve or add/remove entities in the schema

  • Support for Uptraining to add support for unsupported languages

  • Addition of 3 new entity types (traveler_name, reservation_idline_item/transaction_date)

  • Maximum pages (online/synchronous requests) limit has been increased to 15.

Deprecation of the old Stable version

The pretrained-expense-v1.1-2021-04-09 version of the Expense Parser will be deprecated following this release. 

Invoice Parser Updates

The previous Stable invoice processor version, pretrained-invoice-v1.1-2021-04-09, is deprecated as of November 22, 2022.

v1beta3

The Invoice Parser, for v1.3 and v1.4, now has the following quotas and limits:

  • Maximum pages (online/synchronous requests): 15
  • Maximum pages (batch/offline/asynchronous requests): 200

November 16, 2022

v1beta3

The Identity Document Proofing Processor is now available in Public Preview.

The Identity Document Proofing Processor is designed to help predict the validity of ID documents with four different signals:

  • is_identity_document detection: predict whether an image contains a recognized identity document.
  • suspicious_words detection: predict whether words are present that aren't typical on IDs.
  • image_manipulation detection: predict whether the image was altered or tampered via an image editing tool.
  • online_duplicate detection: predict whether the image can be found online.

November 11, 2022

v1beta3

New stable W2 processor version with the following enhancements:

  • Breaks down long entities such as addresses into fine-grained sub-entities: StreetAddressOrPostalBox, AdditionalStreetAddressOrPostalBox, City, State, and Zip. This modification not only improves accuracy, but also entity specificity.
  • Handles wider variations of W2 forms including multi-copies (2,3,4-ups) issued by various payroll vendors.
  • Introduces 8 new entities for Box 12 that represent both codes and values.

New stable Payslip processor version with the following enhancements:

  • Bonus, commissions, holiday, overtime, regular pay and vacation are now part of earning_item/earning_this_period and earning_item/earning_ytd. Captures all types of earnings beyond those categories, and maps them to their respective earning rates, hours and pay (both for the period and year-to-date).
  • Returns year-to-date and current period taxes and deductions.
  • Direct deposits are linked to the corresponding bank account numbers.
  • Returns page numbers, state and federal tax exemptions and filing statuses.

October 31, 2022

v1

A new Release Candidate (RC) version of the Document OCR Processor, pretrained-ocr-v1.1-2022-09-12, is available in the US and EU. This RC can detect document defects.

  • If the document is considered to be defective, the API now returns the same 5 document defect types supported by the Intelligent Document Quality Processor:
    • quality/defect_blurry
    • quality/defect_noisy
    • quality/defect_dark
    • quality/defect_faint
    • quality/defect_text_too_small
  • In addition, it now supports 3 more defect types:
    • quality/defect_document_cutoff
    • quality/defect_text_cutoff
    • quality/defect_glare
  • The defect detection results are in the image_quality_scores field on the Page object in the returned JSON. This additional feature adds latency comparable to OCR processing to the process call.

October 21, 2022

v1

As of October 21, 2022, we have promoted our v1.3 Release Candidate version to a Stable version of the Invoice processor.

Features in the new Stable Invoice processor, version pretrained-invoice-v1.3-2022-07-15.

  • Support for seven new languages: Italian, Portuguese, Romanian, Swedish, Estonian, Latvian, and Lithuanian.

  • Support for uptraining using Document AI Workbench. See Uptrain a specialized processor.

  • Improvements to currency and date normalization.

  • Improvements to line item extraction.

  • Quotas and limits

    • Maximum pages (online/synchronous requests): 15
    • Maximum pages (batch/offline/asynchronous requests): 200
    • Regional availability
      • US (Multi-region), Europe (Multi-region)

Alongside this Stable version, we are launching a Release Candidate version, pretrained-invoice-v1.4-2022-10-21.

October 10, 2022

v1beta3

Known issue (Document Labeling)

  • If you delete one or more documents, and these documents selected for deletion are all associated with an active labeling job, then all documents in that dataset will also be deleted, even if you did not select them for deletion. This is true regardless of the number of documents selected.

    Workaround: Do not delete documents during an active labeling job. You can track active labeling jobs on the Dataset management page, under the category Labeling tasks, located on the right side of the page. If you absolutely must delete documents during an active labeling job, ensure that you also select at least one document that is NOT part of this active labeling job. Then, only the non-associated documents will be deleted, and the remaining documents in the dataset will be preserved.

September 29, 2022

v1beta3

This launch upgrades the lifecycle stage of the Custom Document Extractor (CDE) component of the DocAI Workbench from Private Preview to Public Preview. CDE covers essential workflows for developing custom document extraction processors with E2E UI support:

  • Data import
  • Schema creation and annotation
  • Processor model training
  • Evaluation and troubleshooting
  • Model deployment and version management
  • Human-in-the-loop (HITL) integration for "last-mile" processor quality assurance

Notable new Public Preview Custom Document Extractor (CDE) features include:

  • Progressive data import
  • Direct import of annotated .json files as training or test datasets
  • Data labeling platform integration
  • Dataset export with metadata preserved
  • Auto-labeling using a trained processor version at import to minimize manual annotation efforts
  • Tabular entity end-to-end support
  • Checkbox boolean annotation, training, and extraction
  • Fuzzy matching for more flexible model evaluation
  • Exportable / downloadable model evaluation metrics

The following features have been upgraded:

  • Data import
  • Schema creation and annotation
  • Processor model training
  • Evaluation and troubleshooting
  • Model deployment and version management
  • Human-in-the-loop (HITL) integration for "last-mile" processor quality assurance

Known issues

  • During labeling, checkboxes default to a state that does not reflect the selected or unselected states in the corresponding documents.

    Workaround: Label all checkboxes within a schema, regardless of whether they are selected or unselected,  for optimal checkbox extraction quality and accurate evaluation. When annotating a checkbox by drawing its bounding box, all checkboxes in the CDE annotation user interface will default to an unselected state, and you must manually update the selected or unselected state as necessary.

  • If a processor is deleted when there is an active labeling task ongoing, the task does not stop automatically. Labelers and Labeler Managers will still see the task in their labeling and manager consoles.

    Workaround: Cancel the active labeling task before you delete the processor. Otherwise, navigate to the manager console to pause or delete the corresponding task queue. Any documents labeled from that task after the processor is deleted will be unretrievable.

  • Some discrepancies might exist between pretrained and uptrained versions of a processor.

  • On the Evaluate & Test tab, the evaluation does not show as complete until you click Refresh Table.

  • If you cannot choose a specialist pool due to no chooser for the specialists, clicking Continue causes an error.

    Workaround: Click Task Details to go back and fill in remaining details.

  • Support for nested entities is limited to data arranged in a tabular format. Other layouts are currently not supported.

    Any mention of "nested entities" in previous versions of the Document AI Workbench documentation have been replaced with "tabular entities" to reflect the table-based nesting capabilities for this launch.

  • If you see this message, "This processor does not currently support uptraining or evaluation capabilities" on a processor in the Processor gallery, this means that only the prediction endpoint is available at this time.

  • In some cases, required_once and optional_once entities appear multiple times in a document, with the same value copied to multiple locations. While one annotation is s sufficient when annotating a test set for evaluation, all copies should be annotated to ensure higher recall for trained models.

    Workaround: Annotate all instances of an entity within a doc. This will support both Evaluation and Training without any infrastructure changes.

  • Existing Human in the Loop (HITL) configurations for Label-level filters do not automatically populate to new processor configurations.

    Workaround: If you require changes to the configuration, navigate to the Human-In-The-Loop tab. Under Set filters, select Label-level filters and click Set Label Filters. A Label-level filters page opens. Manually edit this table to configure all of the labels needed for validation and review.

  • Support for handwritten entity detection such as signatures, dates, and initials are limited and may require additional configuration for proper evaluation.

    Workaround: When evaluating pretrained processors, the Value should be manually updated to YES if the entity is present, or NO otherwise. For uptrained processor versions, the Value should be kept as detected by OCR. This issue affects W9, HUD92900B, SSA-1099, and VBA26-0551 processors.

  • When you attempt to select a specialist pool for a labeling task, and there are missing fields in the form, and then click Continue, you might get an error.

    Workaround: Click Task Details, and enter the missing details in the form. You can then successfully create a labeling task.

September 15, 2022

v1beta3

Schema support for checkboxes and nested entitites

  • Customers using Document AI Workbench, and processors for Purchase Order (PO), Invoice, or Expense, now have access to a new schema. This schema enables customers to label checkboxes, if they are defined in the schema, and to accurately represent nested entities, such as parent-child relationships, on the HITL annotation and review console. As additional processors adapt the new schema, these release notes will be updated to include those.

Nested entities

  • The Annotation console now supports labeling for nested entities. The left panel is refreshed with a new look for nested rows to represent nested entities. The value of "parent" will now be the concatenation of all its "children". The parent is effectively a container for all of its children.

September 01, 2022

v1beta3 & v1

We are standarding our release processes and naming conventions for processor versions. For more information, see Manage processor versions.

July 27, 2022

v1

New Release Candidate (RC) versions for PDAI Invoice and Expense processors - July 2022

We have launched new RC versions of Invoice parser and Expense parser on July 15, 2022. These can be accessed in the following way:

  • Invoice parser: pretrained-next-uptrainable
  • Expense parser: pretrained-next

Here are the details about the contents of the RC version updates:

Processor New Languages New Entities
Invoice: pretrained-next-uptrainable Italian, Portuguese, Romanian, Swedish N/A
Expense: pretrained-next Japanese Support for hotel and car rental folios

Payment information entities: Last 4 digits of credit card, payment type

The current limit for uptrainable processors is as follows (it is different from the pre-trained version). We are gathering customer feedback to increase the async limit.

Quotas and limits

Maximum pages (online/synchronous requests): 10
Maximum pages (batch/offline/asynchronous requests): 15

June 30, 2022

v1

VPC Service Control support

Document AI VPC Service Controls provide additional security for your resources and services. To learn more about VPC Service Controls, see the VPC Service Controls overview.

To learn about the limitations when using Document AI with VPC Service Controls, see the supported products and limitations.

June 13, 2022

v1

Document AI is now generally available (GA) in the following new locations:

  • asia-south1 (Mumbai)
  • australia-southeast1 (Sydney)

You must request access to use the new locations. For more information, see Regional and multi-regional support.

v1beta3

New Identity Processor (Preview)

The France Passport Parser is now available in limited preview.

June 10, 2022

v1beta3

The Contract Parser is now more accurate, can extract more fields and supports higher page limits.

June 01, 2022

v1

Identity DocAI General availability (GA) release

The following Identity DocAI processors are now Generally Available (GA).

For more information, see Document AI for Identity.

April 21, 2022

v1

Document OCR processor

The changes from the Google Default Next version have been applied to the Google default version.

The previous Google default version can still be accessed until July 21, 2022 as pretrained-legacy. After July 21, 2022, that version will be removed.

For more information about using different versions of the processor, see Managing processor versions .

For the original announcement of this change, see the January 14, 2022 release note.

April 08, 2022

v1

New Version of Lending W2 Processor

We have released a new Release Candidate version of the W2 Processor. This version is experimental and has the following features:

  • Quality improvement on SSN and EIN fields.
  • Support for box 12 fields, including both codes and values.
  • Fine grained predictions of EmployeeName, EmployeeAddress, and EmployerNameAndAddress which are no longer part of the output and replaced with additional fields.

March 25, 2022

v1

New & Updated processors available

The following Lending DocAI processors are now available for trusted testers. Access to the trusted testers program is limited and granted on a case by case basis. If you would like to be considered please fill out the DocAI Processor Access Request Form:

New Experimental processors to support new document types:

  • Form VA Loan Discharge Statement Processor
  • Form USDA Conditional Statement Processor
  • Form 1017 Processor
  • Form Biweekly Payment Rider Processor
  • Form VBA26 1805 Processor
  • Form VBA26 6393 Processor
  • Form MERS Rider Processor

Updated Experimental processors:

  • Form 4506-T Processor
  • Form 4506-C Processor
  • Form HUD54114 Processor
  • Form HUD92900WS Processor
  • Form HUD92800 Processor
  • Form 1040-NR Processor
  • Form HUD92900LT Processor
  • Form VBA26 8923 Processor
  • Form HUD92900A Processor
  • FORM_1005_PROCESSOR

March 10, 2022

v1

Document AI is now generally available (GA) in the following new locations:

  • europe-west3
  • asia-southeast1

You must request access to use the new locations. For more information, see Regional and multi-regional support.

February 18, 2022

v1

New Versions of Procurement Processors

We have launched a new Google Pretrained version of the following procurement processors with various quality improvements:

The changes from the old Google default next version have been applied to the new Google Pretrained version. The old Google default version is still available and will not be deprecated for at least 180 days.

January 26, 2022

v1beta3 & v1

Enrichment using the Knowledge Graph is now Generally Available.

For more information, see Enterprise Knowledge Graph field enrichment.

January 14, 2022

v1

Document OCR processor

We have updated the Google default next version with quality improvements. Consequently, you have 90 days from today to test the new model before the changes are applied to the Google default version. After that, the original Google default version will be available for another 90 days as legacy. For more information about the processor and its versions, see the Document OCR processor.

For more information about using different versions of the processor, see Managing processor versions.

For the original announcement of this change, see the November 5, 2021 release note.

December 15, 2021

v1beta3 & v1

New Lending Processors (Preview)

The following new processors are now available in limited preview:

New Versions of Lending Processors

We have launched new versions of the following lending processors.

These new versions use a new lending document splitting and classification model with improved quality and support for more document types. For more information, see the Document types identified by the Lending Splitter & Classifier.

November 10, 2021

v1

We have lowered the price for many processors. For more information, see the Pricing page.

November 05, 2021

v1

The following procurement processors are now publicly accessible:

We have release a new version of the Document OCR Processor called Google default next. This version changes the distribution of confidence scores in the response. You have 90 days from today to test the new model before the changes are applied to the Google default version . After that event, the original version will still be available for another 90 days as legacy. For more information about using different versions of the processor, see Managing processor versions.

New Lending Processor (Preview)

The Mortgage statement parser is now available in limited preview.

October 15, 2021

v1beta3

Contract DocAI (Preview) released

The Contract parser is now available.

October 06, 2021

v1

Document AI is now generally available (GA) in the following new locations:

  • europe-west2
  • northamerica-northeast1

You must request access to use the new locations. For more information, see Regional and multi-regional support.

September 01, 2021

v1

Document AI now supports Data Residency, VPC-SC, Access Transparency, and CMEK.

August 20, 2021

v1

Managing processor versions

You can now switch between different versions of a processor. For more information, see Managing processor versions.

New processor versions

We have added new versions of the following processors:

  • Bank statement parser: improved model quality
  • Pay slip parser: improved model quality and extraction of three additional fields: net_pay, net_pay_ytd, and employee_account_number.

New Lending DocAI processors

The following Lending DocAI (LDAI) processors are now available in limited Preview:

  • 1065 parser
  • 1099-NEC parser
  • 1099-R parser
  • 1120 parser
  • 1120-S parser
  • SSA-1099 parser

Additionally, the LDAI Document Splitter and Classifier has been updated to support the new LDAI processors as well as the following processors:

  • US Driver License Parser
  • US Passport Parser

Human in the Loop (HITL) support for Lending DocAI processors

The following Lending DocAI processors now support Human in the Loop (HITL):

  • 1003 parser
  • 1040 Parser
  • 1040 Schedule C parser
  • 1040 Schedule E parser
  • 1099-DIV parser
  • 1099-G parser
  • 1099-INT parser
  • 1099-MISC parser
  • Bank Statement parser
  • Pay Stub parser
  • W2 parser
  • W9 parser

Knowledge Graph support

The following processors now support Knowledge Graph enrichment:

  • Bank Statement
  • Pay Slip
  • W2 Parser
  • W9 Parser

July 30, 2021

v1

The Invoice Parser now extracts a new field invoice_type that indicates the type of the input document.

July 02, 2021

v1

Change in processor documentation

The location of individual processor information has changed. You can now find individual processor documentation for all solutions (General, Procurement, Lending) in the following locations:

Human in the Loop (HITL) now supports priority queues for each processor, based on the urgency of each document. For more information, see HITL.

June 09, 2021

v1

VPC Service Controls

Integration with Document AI VPC Service Controls is now generally available.

April 09, 2021

v1

Procurement DocAI General availability (GA) release

Procurement DocAI (PDAI) solution is now available in private General Availability (GA).

This includes the following processors:

Human in the Loop (HITL) support for Procurement DocAI processors

Procurement DocAI processors now support Human in the Loop (HITL) AI platform functionality supporting human revisions of predictions.

Invoice parser behavior update

The invoice parser behavior has been updated to include the following features:

  • Offers extended support for the following languages (in addition to English):
    • French
    • Dutch
    • German
    • Spanish
  • Improves supplier parsing accuracy with Knowledge Graph support.
  • Improves prediction quality (accuracy).
  • Extends the header and line item fields extracted by the parser.
  • Increased the number of pages for online processing (10 pages) and offline processing (200 pages).
  • Increased the number of documents per batch in offline processing (50 documents).

Expense parser (Receipt parser) behavior update

The expense parser behavior has been updated to include the following features:

  • Renamed Receipt parser to Expense parser.
  • Improved prediction quality.
  • Improved prediction quality for English, French, and Dutch for more expense types (for example hotel statements).

Human in the Loop (HITL) AI General Availability (GA) released

HITL AI is now available in Private General Availability (GA) for human review of Invoice, Expense, and Utility parser predictions.

Features:

  • HITL configuration enhanced to designate which fields need review and whether a field is mandatory, saving review time.
  • Labeler UI highlights the fields below a confidence score and supports single-click confirmation to improve review efficiency.
  • Labeling Manager shows analytics and metrics by task and by labeler to streamline HITL operations.

April 02, 2021

v1

Lending DocAI General Availability (GA) released

Lending DocAI is now General Availability. See the documentation for more information.

Lending DocAI processors added

The following Lending DocAI processors are now available:

March 31, 2021

v1

Document AI General availability (GA) released

Document AI is now General Availability (GA).

January 14, 2021

v1beta3

New Procurement DocAI processor released in limited Preview

The following Procurement DocAI processor is now available in limited Preview:

  • Procurement document splitter

For more information, see the processor documentation.

January 11, 2021

v1beta3

Lending processors behavior update

The behavior of the following processors has been updated:

  • 1003 parser
  • 1040 parser
  • 1099-MISC parser
  • W2 parser
  • W9 parser

Now, if these processors are given a multi-page input file and contains a page that is the correct document type and one of the supported versions the processor performs entity extraction for that page; subsequent applicable pages will not be processed. If the prcoessor doesn't find any applicable documents in the input file it returns an error message.

October 29, 2020

v1beta3

Document AI Preview released

The following beta and preview features are available in API version v1beta3:

  • Procurement DocAI processors: Invoice parser and receipt parser.

October 16, 2020

v1beta3

Document AI Preview released

The following beta and preview features are available in API version v1beta3:

  • General processors: Document OCR (Optical Character Recognition), form parser, and document splitter.
  • Lending processors: W9, 1040, W2, 1099-MISC, and 1003 parsers, as well as lending document splitter & classifier.

uri field unavailable

  • Sending a request with the uri field is currently not supported for v1beta3. Any updates to the availability of the uri field will be announced here.

Workaround: Send requests with image information in the content field (base64 encoded information).

August 24, 2020

v1beta2

Form Parser model updates

The Form Parser model has been updated. The model update includes the following features:

  • Improved OCR quality for English detection.
  • Improved key-value pair, checkbox, and table parsing detection quality, particularly for rotated images and handwritten text.
  • Decreased latency for complex tables.

August 20, 2020

v1beta2

Invoice Parsing updates

  • Document AI now supports normalized values for certain entities returned from Invoice Parsing requests.
  • We have improved confidence scores for entities returned from Invoice Parsing requests.

July 04, 2020

v1beta2

Invoice Parsing Beta model upgrade

The Invoice Parsing Beta model has been upgraded. This model upgrade results in higher quality results for the entities and entityRelations. There is no API change.

See the product documentation for more information.

April 14, 2020

v1beta2

Document AI Beta released

The following beta features are available in API version v1beta2:

  • Document processing: You can use the API to parse forms or tables from PDF, TIFF, or GIF documents.
  • Regional support: The API now offers multi-regional support (us and eu) for all features. Using a multi-region endpoint enables you to configure the API to store and process your data in the United States or European Union.

Invoice processing Beta

  • Invoice processing is now available as a restricted feature. See Parsing invoices for more information.