Document data

Document Data is a resource that serves both as an input and output variable, within your Document Understanding workflows. The Document Data object holds all the necessary information about a single document. If you classify a document, the object includes the Document Type. If you extract data, the object contains the corresponding extracted fields. Irrespective of the activity, Document Data consistently contains the document's text and DOM (Document Object Model).

With Document Data you can: collect all the necessary information about a document in one variable, save data to each property of the object, and reuse it for other activities in the workflow.

Document Data holds information about the following attributes:

• DocumentType: Provides data about the identified Document type, populated by activities such as Classify Document or Create Classification Validation Task.
• Data: Contains the extracted field values. It is generated on demand by the Generate Data property, which generates an output type of IDocumentData<ExtractorType>. When the Generate Data property is set to False, you can access the extracted field values only through methods of type Get.
• FileDetails: Contains details about the IResource.
• SubDocuments: Includes a collection of Document Data, populated by activities such as Create Classification Validation Task.
• DocumentMetadata: Contains information about processing the document, such as:
  • Text detected language
  • Extracted fields as Data Table
  • Document Object Model (DOM): Holds the Document Object Model which is used by all activities.
  The DocumentMetadata is updated by the activity that first processes the document. Once populated, the metadata is shared with and used by all the succeeding activities that receive the Document Data object.
  Tip: Unless an activity is the first Document Understanding activity part of a Studio workflow, use Document Data as input. Use the File variable as input only if the activity is the first Document Understanding one part of a Studio workflow.

Understanding and accessing Data

The Document Data contains detailed information about extraction results. The Data can be of two types, depending if you choose to generate it on demand, so you can further use it in your workflow, or to not generate it, so you can perform get and set methods on it, designed for advanced implementations, to increase flexibility.

The Generate Data Type property in the Extract Document Data activity allows you to choose if you generate the data on demand or not. Refer to the following scenarios:

• When you set Generate Data Type to True (the default setting): Document Data outputs as IDocumentData<ExtractorType>. This data is generated on demand and changes based on modifications made in the Extract Document Data activity. With this setting, you cannot change the document type in the Validation Station, and JIT (Just in Time) is selected by default.
• When you set Generate Data Type to False: Document Data outputs as IDocumentData<DictionaryData>. With this setting, the Document Data property will not be generated anymore, and you won't be able to browse through it.

  You can access its data using specific methods, relying on the field ID. These IDs become available when configuring the document type or when retrieving the information using APIs. Visit Editing or adding new fields and Get the extraction request API for more information.

  • When you set Generate Data Type to False for generative extraction, the retrieved fields correspond to the names provided in the prompt. For example, if the field name in the prompt is defined as a b c (including the spaces), you should use the same as the field ID when using the specific methods.

Passing Document Data to activities

When you use Document Data, the first output object is made from your input file. After you created this object, we recommend you to pass it along to your next activities. By passing it along to your next activities, you can reuse the Text and DOM from your original file. This approach saves you from re-digitizing the file each time.

Consuming the extraction results for single and multi-value fields

If you configure a document type field to be multi-valued, the system expects multiple values. An example might be a multiple-choice question on a form. The results appear in the multi-value attribute on the field, returned as a list. If the document type field is configured to be single value, the system returns the result in the value attribute on the field by default.

The following table shows you how Document Data returns single and multi-value fields:

Table 1. The values returned by Document Data for single and multi-value fields

Field type	Has no value	Has one value	Has two or more values	DocumentData.Data.FieldName.Value	DocumentData.Data.FieldName.MultiValues
Single value	Yes	No	N/A	""	null
Single value	No	Yes	N/A	<value that was identified>	null
Multi-value	Yes	No	No	""	[] (empty array)
Multi-value	No	Yes	No	<value that was identified>	[<array with one value identical to the .Value>]
Multi-value	No	No	Yes	<first value that was identified>	[<array with n values, with the first value being identical to the .Value>]

Returning extracted fields as a Data Table

You can return the fields you extracted from a document as a Data Table, using the Document Data object. You can then use the Data Table variable inside Excel activities.

To return the extracted fields as a Data Table, choose the ResultsAsDatatable output for the Extract Document Data activity.

Properties

The properties of the Document Data variable can be populated and consumed by one or multiple activities. Depending on the activity populating the variable, the properties can differ. Check the following list:

• DocumentType - Classify Document activity populates the following values:
  • DisplayName (used for custom models): Name of the Document Type.
  • ID (used for out-of-the-box models): Name of the Document Type.
  • Confidence: Classification confidence.
  • URL: URL of where the Document Type is accessible; this can be either custom or predefined, referenced via the respective project in Document Understanding center.
• Fields - Extract Document Data, Create Validation Task, Create Validation Task and Wait, Wait for Validation Task and Resume activities populate the following values:
  • Field Value: Extraction value of the field.
  • Extraction Confidence Score: Confidence score of the extraction, as provided by the model.
  • OCR Confidence Score: Confidence score provided by the OCR engine.
• File Details - Activities creating the Document Data object, receiving a file as input, populate the following values:
  • Full Name: Full name of the file.
  • Extension: Extension of the file.
  • Page Range: Page range of the file.
• Sub Documents: Collection of Document Data, populated by the Classify Document activity.
  Note: This is not currently populated and will be added in the future together with classification validation and splitting capabilities.
• DocumentMetaData:
  • DOM: The document object model, used by all activities. (populated by activities creating the Document Data object, receiving a file as input.)
  • Text: All extracted text. (populated by activities creating the Document Data object, receiving a file as input.)
  • Language: The language detected in the document. (populated by activities creating the Document Data object, receiving a file as input.)
  • Split confidence: If the document is split, the document is returned by the splitting model. (populated by the Classify Document activity)
    Note: This is not currently populated and will be added in the future together with classification validation and splitting capabilities.
  • Results as Data Tables: Fields exported as Data Table. (populated by the Extract Document Data activity).