Field Identification tasks are used for Semi-structured documents. In these tasks, you confirm or correct predicted field locations and identify fields that the system could not locate. A task may be created when the system is uncertain about a field location or when a field is configured for manual identification in the Layout Editor.
Accessing this feature
Your access to the feature described in this article depends on your license package and pricing plan.
To learn which features are available to your organization and how to add more, contact your Hyperscience representative.
Start a Field Identification task
Go to Tasks > Perform Tasks.
In the Supervision Tasks section, find Field Identification, and then click Perform Tasks.
Review the task guidance, and then continue to the task. You can choose not to show the guidance again.
Before you begin, review Best Practices for Field ID Supervision and QA.
Field Identification interface
The task interface has three main areas:
Page thumbnails on the left let you move between pages and reorder pages when necessary.
Document view in the center shows the active page, text segments, predicted locations, and the bounding boxes you create.
Locate Fields on the right lists Fields to Locate and Pre-identified Fields. It also contains Document details and the action for completing the task.
Identify fields
Select a field in the right panel. If a prediction is available, the predicted location appears as a blue bounding box. If no prediction is available, locate the field in the document.
Select the field that you want to identify.
Click the text segment containing the field value. A bounding box appears around the selected segment.
Review the bounding box and adjust it when necessary.
Continue until you have reviewed every field. If a field is not present, leave it without a bounding box.
If... | Then... |
|---|---|
The bounding box contains the complete field value. | Continue to the next field. |
The box is in the correct location but cuts off part of the value. | Resize the box until it contains the complete value. |
Adjacent text segments are part of the same field value. | Draw a box around all of the segments that belong to the value. |
The box is in the wrong location, or no box appears when you select the value. | Hold the spacebar and draw a box around the correct value. |
Documents with tables
When a document contains fields and tables, complete Field Identification first. The task then proceeds to Table Identification. For instructions, see Table Identification.
Identify multiple occurrences of a field
An occurrence is a distinct value for the same field. For example, a joint account may contain two different values for an Account Owner field. Use multiple occurrences for distinct values that must be extracted. If the same value appears more than once, identify only the first instance in the natural reading order.
Multiple occurrences must be enabled for the field in the Layout Editor. The Field Identification model can also predict multiple occurrences. For configuration and training information, see Training an Identification Model.
Identify the first occurrence.
In the right panel, click Add another for the field.
Identify the next occurrence.
Repeat these steps for every additional distinct value.
Remove an occurrence
Use the Delete icon beside an additional occurrence to remove it. If the occurrence contains multiple bounding boxes, removing the occurrence clears all of its bounding boxes.
By default, the next occurrence becomes active and the document may move to the page containing it. To remove an occurrence and remain on the current page:
Hold Shift while clicking the Delete icon.
When the updated keyboard shortcuts are enabled, press Option + Shift + X on Mac or Alt + Shift + X on Windows.
Identify values with multiple text segments
For text values that span lines or pages and cannot be captured in one bounding box, add multiple text segments to the same occurrence.
Select the field and identify the first text segment.
Click Add another text segment.
Identify the next segment.
Repeat until every segment belonging to the value is identified.
Additional text segments are supported only for text fields.
Search for text segments
Use Search segment text to locate text across all pages of the document. Search matches are not case-sensitive and can match part or all of a segment’s transcription.
Enter the text in the search field, and then press Return on Mac or Enter on Windows. You can also focus the search field by pressing Command + Option + F on Mac or Control + Alt + F on Windows.
Use Previous Segment and Next Segment to move through the results in reading order. The result counter shows your current position.
Clear the search to remove the search highlights.
The active result is highlighted in dark orange; the other matching segments are highlighted in light orange.
Work with multi-page documents
Select a thumbnail in the left panel to move to another page. To reorder pages, drag a page thumbnail to the required position.
Review all pages before deciding that a field is not present. When a field value spans pages, use multiple text segments for the same occurrence.
Handle an incorrect layout
If the document was matched to the wrong layout and reprocessing is available:
Expand Document details in the right panel.
Click Mark Layout Variation Incorrect.
Confirm the action by clicking Yes, Mark Incorrect.
The document’s pages are sent for reprocessing. For more information, see Document Classification Task.
Complete the task
After you have reviewed the fields, click Confirm All and Continue. If an additional occurrence has no location, the Empty Occurrences message lists the affected fields. Identify a location for each occurrence or delete the empty occurrence before continuing.
Long-form Extraction
Long-form Extraction uses the UNSTRUCTURED_EXTRACTION Field Identification model type to locate long text values in unstructured documents. The Field Identification task uses the same review workflow: select the field, identify its complete value, and adjust the bounding box when necessary.
For setup, eligibility, and model-training information, see Training an Identification Model.
Keyboard shortcuts
Press F2 in the task to open the keyboard shortcuts panel. The panel shows the shortcuts that apply to the current task and operating system. If multiple occurrences or multiple text segments are not available for the task, their shortcuts are not shown.
The updated shortcuts are optional in v43.2 and can be enabled or disabled from the keyboard shortcuts panel. Learn more in Keyboard Shortcuts.