UiPath Documentation
activities
latest
false
重要 :
请注意，此内容已使用机器翻译进行了部分本地化。 新发布内容的本地化可能需要 1-2 周的时间才能完成。
UiPath logo, featuring letters U and I in white

Document Understanding 活动

上次更新日期 2026年4月22日

发行说明

3.1.0 - preview​

发布日期：2026 年 4 月 1 日

改进​

增强型遥测和消耗跟踪

Document Understanding activities now include improved telemetry and consumption tracking capabilities.

3.0.0 - 预览​

发布日期：2026 年 2 月 25 日

重大变更​

  • 已删除对“Windows - 旧版”项目的支持。从此版本开始，包仅与 Windows 项目兼容。
  • Starting with this version, upgrading the package requires verifying that any downstream processing using the output of the Extract Document Data activity is still functioning correctly. The structure of the data returned by IXP projects has changed significantly. Therefore, it is strongly recommended to validate that any post-processing logic remains compatible. What has changed:
    • Results are now reported as Field Groups instead of Tables.
    • 现在，值可以正确反映非结构化文档 IXP 项目中定义的字段类型。
    • Monetary Quantity is now returned as a dedicated field type.
    • Document Validation tasks automatically display results as Field Groups instead of Tables.
    • Data manipulation logic must be updated to handle Field Group values accordingly.

新增功能​

To provide extraction capabilities tailored to a specific non-Latin document layout, you can now use a dedicated modern project type called Predefined Non-Latin Languages in the Extract Document Data activity. This project type offers the following extractors:

  • Invoices Japan – Recommended for Japanse invoice documents. The extractor can handle common Japanese invoice layouts, and can identify and extract key invoice fields such as supplier information, invoice number, and currency.
  • Invoices China - Recommended for Chinese invoice documents. The extractor can handle common Chinese invoice layouts, and can identify and extract key invoice fields such as supplier information, invoice number, and currency.
  • Receipts Japan - Recommended for Japanese receipt documents. You can use the extractor to identify and extract fields such as merchant name, transaction date, total amount, tax, and currency from Japanese-language receipts.

For more information visit Extract Document Data.

2.17.4​

发布日期：2026 年 3 月 16 日

错误修复​

Fixed an issue where the Classify Document and Extract Document Data activities returned a resource not found error after migrating a DU project. Even when you configured the activities to use a tag, they internally used version numbers to locate classifiers and extractors. Because version numbers reset during migration, the activities could not find the correct resources. The activities now use the configured tag to locate resources. As part of this fix, the ProjectVersionName property is no longer visible in Studio for these activities.

已知问题​

After upgrading to UiPath.DocumentUnderstanding.Activities version 2.17.0 or 2.17.4, if Generate Data Type is enabled in Studio Desktop, the Extract Document Data activity may throw the following error: <type> is ambiguous in the namespace <namespace>.

此问题将在即将发布的 Studio 桌面版中解决。

2.17.0（最新版本）​

发布日期：2025 年 10 月 28 日

新增功能​

在分类期间对文档中的页面重新排序

For greater flexibility, you can now reorder pages within a document, when using the Classification Station. To enable this feature, use the EnablePageReordering property, available in the Create Classification Validation Task and Create Classification Validation Task and Wait activities.

验证站点精简模式

Validation Station can now be used with a new face: the Compact Display Mode. This provides more space for fields, easier highlighting for the fields that require attention, a centralized business checks section, and a host of other validator-centric features. To enable the Compact Mode, go to the Create Validation Task or Create Validation Task and Wait activities, and set the DisplayMode input argument to Compact.

错误修复​

  • 修复了在“提取文档数据”活动中选择资产会导致设计时错误的问题
  • 修复了与通过分类站点处理文件以及对子文档进行后续数据提取操作相关的许多问题。

已知问题​

After upgrading to UiPath.DocumentUnderstanding.Activities version 2.17.0 or 2.17.4, if Generate Data Type is enabled in Studio Desktop, the Extract Document Data activity may throw the following error: <type> is ambiguous in the namespace <namespace>.

此问题将在即将发布的 Studio 桌面版中解决。

Erratum April 16, 2026: Due to a mishap on our side, this known issue was not made available on the original announcement day.

2.16.1​

发布日期：2025 年 9 月 11 日

错误修复​

修复了导致 Action Center 中的某些操作无效的错误。仅当使用生成式提取程序或外部应用程序配置了提取文档数据活动时，才会发生此问题。

2.16.0​

发布日期：2025 年 9 月 11 日

错误修复​

我们修复了以下问题：在使用外部应用程序配置在 Document Understanding 新式项目中配置活动时，Studio 显示与租户相关的错误。

已知问题​

Actions in Action Center become invalid when the Extract Document Data activity is configured with either the Generative Extractor or an external application. This issue was fixed in version 2.16.1.

v2.14.0​

发行日期：2025 年 4 月 16 日

按分类的文档类型提取文档数据：根据分类操作的结果使用提取程序​

In your classic and modern projects you can now use the output of the Classify Document activity within the Extract Document Data activity. This allows you to automatically select the appropriate extractor for processing by identifying the extractor that can process the documents of the type provided for DocumentData.DataType. In the list of extractors, choose Use Classification Result for the suggested extractor. Visit Extract Document Data for more information about using classification results to select an extractor.

新增功能​

使用新的提取程序增强了提取功能​

To provide extraction capabilities tailored to a category of documents, you can now use a dedicated modern project type called Generative Predefined. This project type offers the following extractors:

  • Long Documents Simple Layout – This is the existent Generative Extractor.
  • Long Documents Complex Layout (Preview) – Optimized for long form documents that include images, handwriting, form elements or other complex layouts, such as floating callout boxes. Examples of documents that are suitable for this extractor: insurance policies, or other similar long form documents with complex layouts.
  • Short Document Complex Layout (Preview) – Optimized for short documents that include images, handwriting, form elements or other complex layouts like floating callout boxes. For example: identity cards, or healthcare intake forms with complex layouts.

For more information visit Extract Document Data.

使用标签和版本引用 Document Understanding 新式项目​

To consume a snapshot of a model, activities and APIs now allow you to consume certain versions of your projects during classification and extraction. The Tag and Version properties provide granular control for consuming a version of a published model. For more information about consuming versions with DocumentUnderstanding.Activities, visit Classify Document and Extract Document Data. For information about exposing the version in your project, visit Document Details. For information on the newly available APIs for consuming tags and versions, visit Discovery APIs and Digitization APIs.

支持本地部署设置中的活动​

You can now use the Classify Document and Extract Document Data activities even if the robot is connected to a local Orchestrator. At design-time, you can use Document Understanding resources from different organizations or tenants. Similarly, at runtime, you can execute these activities while connected to a local Orchestrator in Studio.

The new Design-time external connection and Runtime external connection properties allow you to directly use external application credentials, or credentials stored in Orchestrator, to access Document Understanding resources during design-time or runtime.

备注：

Ensure your selected tenant has Document Understanding enabled. If you create a runtime external connection, then make sure that you also allocate AI Units to the selected tenant.

改进​

  • For the Generative Predefined project, the existent Generative Extractor is now called Long Document Simple Layout Extractor.
  • When selecting an extractor for the Generative project types, the former Prompt collection in the Classify Document and Extract Document Data activities, has been renamed to Document Type details, which describes the purpose of the collection more accurately.
    • Inside the updated Document Type details collection, the former Generative prompt column is now Instruction. The Instruction field represents the instructions you want to offer about the information that should be extracted for a certain field name.For more information about the updates, check the Classify Document and Extract Document Data activities.
  • 文档处理工作流的数字化部分现在可以识别并返回包含带引号圈出的值、带下划线的值或划掉的值的文档中的数据。

v2.12.1​

发布日期：2024 年 11 月 19 日

错误修复​

由于默认超时，长时间运行的分类和提取流程失败。我们现在考虑将活动中设置的所有超时的总和作为所有后台操作的全局超时。此外，我们改进了错误消息，以便在此特定情况下更好地明确失败原因。

v2.4.5​

发布日期：2024 年 10 月 21 日

错误修复​

This version resolves the Unknown sections in the AI Units consumption overview dashboard that were caused by missing data.

v2.9.6​

发布日期：2024 年 10 月 14 日

错误修复​

This version resolves the Unknown sections in the AI Units consumption overview dashboard that were caused by missing data.

v2.12.0​

发布日期：2024 年 10 月 3 日

新增功能​

禁用按需生成文档数据，以增强高级实施的灵活性​

You can now change the output type of the Extract Document Data activity using the new Generate Data Type property. When you set this property to False, the output type changes from IDocumentData<ExtractorType> to IDocumentData<DictionaryData>. Setting Generate Data Type to Flase simplifies retrieving and changing field values, and allows you to change the document type in the Validation Station.

Visit the Extract Document Data and Document data pages to check how to use the Generate Data Type property and what methods you can use to access the extraction results.

改进​

You can now set the Orchestrator storage bucket to work with Additional options for the Create Classification Validation Task and Create Classification Validation Task and Wait activities. If there is no specific storage bucket created, you can create a default one.

错误修复​

  • 在 Mac 机器人上运行 Studio Web 工作流失败，出现以下异常：“无法加载文件或程序集‘UiPath.DocumentUnderstanding.Common.SDK’”。
  • 手动更正文档中的某些数字或日期时，负责创建 Action Center 任务的活动发生崩溃。
  • 修复了恢复工作流时 Studio Web 中发生的错误。以前，在首次验证文档分类后会发生此错误，其中涉及从 Microsoft OneDrive 检索文件并在 Action Center 中对其进行验证。

v2.2.6​

发布日期：2024 年 10 月 3 日

我们通过将常见依赖项更新为最新版本来提高产品稳定性。 此升级是自动的，不需要您执行任何操作。

v2.9.5​

发行日期：2024 年 7 月 31 日

错误修复​

When attempting to load a document attached to an entity in UiPath® Apps, workflows in Studio Web that use the Extract Document Data activity no longer result in a "System.InvalidOperationException" error.

v2.9.4​

发行日期：2024 年 7 月 12 日

错误修复​

修复了导致文档分类失败并显示“提供了无效的请求 URI”错误的问题。

v2.9.3​

发行日期：2024 年 6 月 20 日

错误修复​

手动更正文档中的某些数字或日期时，负责创建 Action Center 任务的活动发生崩溃。

v2.4.3​

发行日期：2024 年 6 月 5 日

我们通过将常见依赖项更新为最新版本来提高产品稳定性。 此升级是自动的，不需要您执行任何操作。

v2.9.2​

发行日期：2024 年 6 月 5 日

我们通过修改某些依赖项提高了产品稳定性。

v2.9.1​

发行日期：2024 年 5 月 27 日

改进​

  • 提示大小从每个问题 500 个字符增加到 1000 个字符，以提高说明的清晰度。 此外，如果达到每个问题 1000 个字符的提示大小限制，您将收到“超出限制”错误。
  • 增强了“提取文档数据”活动的错误消息，使其清晰度更高，调试更容易。

错误修复​

  • The document type remained as Unknown in the Validation Station after using the Generative Extractor and Generative Classifier activities, or after manually assigning the document type.

v2.9.0​

发布日期：2024 年 4 月 29 日

这些发行说明包含 2023 年 11 月至 2024 年 3 月之间进行的所有更新。

备注：

In case of upgrading from version earlier than v.2.4.1, make sure to reopen the workflow after upgrading the UiPath.DocumentUnderstanding.Activities package.

新增功能​

提高提取文档数据的提取置信度

We've improved our Extract Document Data activity to increase score accuracy and decrease validation time. This enhancement the Auto-validation and Confidence threshold properties, enabling the cross-verification of extraction results from specific models against a generative model.

Visit Extract Document Data to learn how to increase your extraction confidence levels.

分类验证任务

我们很高兴地宣布，您现在可以参加以下活动：

改进​

  • You can now set the Orchestrator storage bucket to work with Additional options for the Create Validation Task and Wait and Create Validation Task activities. If there is no specific storage bucket created, you can create a default one.
  • The optional Timeout property is now available for the Extract Document Data and Classify Document activities. This parameter configures a timeout for the activities.
  • The selected extractor in the Extract Document Data activity now overrides the document type. Doesn't apply to generative models.
  • In case of multi-value fields, all values are returned under Document Data for the Extract Document Data activity. The values are available in DocumentData.Data.FieldName.MultiValues[].
  • This release brings the following updates to the Document Data object:
    • The Name property from the Document Type attribute is replaced with the following:
      • DisplayName for custom models
      • ID for out-of-the-box models
    • 添加了两个新属性，并根据 Document Understanding 框架的结果填充：
      • ID
      • 显示名称

已知限制​

当前处于预览阶段的现有 Document Understanding Insights 仪表板不再显示跨平台 DocumentUnderstanding.Activities 包中的数据。 现在，它们仅报告来自 IntelligentOCR.Activities 工作流的数据。

来自跨平台 DocumentUnderstanding.Activities 的数据现在在单独的新 Insights 仪表板中报告。

错误修复​

  • The ClassificationResults output property of the Create Classification Validation Task activity is renamed to DocumentData. CAUTION:
    备注：

    如果CreatedClassificationValidationTask.ClassificationResults属性在当前工作流中，则在升级后将无法使用该属性。

  • The output property ExtractionResults of the Create Validation Task activity is renamed to DocumentData. CAUTION:
    备注：

    如果您的工作流使用CreatedDocumentValidationTask.ExtractionResults属性，则在升级后将无法使用此属性。

  • Fixed an issue where the Wait for Validation Task and Resume activity didn't recognize numbers in the 3.1342,7 format, causing an "Input string was not in correct format" error. All number formats are now fully supported.

v2.4.2​

发行日期：2024 年 4 月 10 日

错误修复​

Resolved an issue where the Extract Document Data activity opened with an error that could not be reviewed because the activity didn't expand.

v2.4.1​

发布日期：2023 年 11 月 1 日

生成式功能正式发布​

我们很高兴地宣布，UiPath.DocumentUnderstanding.Activities 生成式功能现已正式发布。其中包括以下各项：

  • “分类文档”活动
  • 提取文档数据活动

新特性和改进功能​

macOS 系统现在支持 UiPath.DocumentUnderstanding.Activities 包。

A new property, Extracted fields validation confidence %, is added for the following activities:

  • 创建验证任务并等待
  • 等待文档验证任务然后继续

v2.4.0​

发布日期：2023 年 10 月 11 日

Document Understanding 活动正式发布​

我们很高兴地宣布，UiPath.DocumentUnderstanding.Activities 包现已正式发布。

备注：

Generative features are still in public preview.

v2.2.4​

发布日期：2023 年 9 月 28 日

新特性和改进功能​

macOS 系统现在支持 UiPath.DocumentUnderstanding.Activities 包。

v2.2.3​

发行日期：2023 年 5 月 11 日

错误修复​

  • 我们修复了在使用模板时导致“提取文档数据”活动停止加载的错误。
  • We fixed a bug where ActionCatalog fields wouldn't work for Create Document Validation Action and Create Validation Task and Wait activities.
  • We fixed a bug that was causing an error when Classify Document activity was used in a workflow.

v2.2.2​

发行日期：2023 年 5 月 11 日

错误修复​

我们修复了一个错误，即用户在尝试使用通过“提取文档数据”活动提取的字段的值并且该值不存在时会收到错误。

已知问题​

macOS 系统不支持 UiPath.DocumentUnderstanding.Activities 包。

v2.2.1​

发行日期：2023 年 5 月 5 日

新特性和改进功能​

The Extract PDF Text activity is the latest addition to the package, allowing you to extract all characters from a specified PDF file and store it in a string variable. When the Apply OCR option is enabled, it extracts the information using OCR, and when disabled it extracts the native content.

已更新两个活动的名称，并已更新一个活动的字段：

  • Create Document Validation Task became Create Validation Task.
  • Wait for Document Validation Task and Resume became Wait for Validation Task and Resume.
  • Set PDF Password now offers more detailed field names, such as:
    • 新的管理密码
    • 新的打开密码
    • 当前管理密码
    • 当前打开密码

此页面有帮助吗？

连接

需要帮助? 支持

想要了解详细内容？ UiPath Academy

有问题? UiPath 论坛

保持更新