Human-in-the-Loop Feature

Overview

On the data review page, the user can view the extracted data that are received from the engagement. It is a Human-in-the-Loop (HITL) feature that allows users to review and edit the extracted data. It also allows the user to add additional data that is relevant to the user before moving to the next stage.

Data Review Page

Follow the below steps to navigate the data review page:

1. The Workflow drop-down menu displays the workflow of the displayed engagement. You can navigate back to the Data Validation page by selecting the applicable workflow.

   

2. You can access other engagements initiated within the same workflow from the drop-down menu.

   

3. The review page has a two-column layout. The left panel displays extracted data with corresponding labels, and the right panel shows the source of the extracted data such as email and PDF attachments (native PDFs and scanned PDFs). The left panel displays three tabs:

   

   • Extracted Data tab: Select this tab to display the extracted data from the source which is email and the pdf attachments. Users can locate the data on the source just by clicking the appropriate field.

     

   • Data in Review tab: Select this tab to display the data marked with a low confidence badge, indicating the need for human review before proceeding.

   • All Data tab: Select this tab to view all extracted data and optional fields with predefined labels.

4. The below-mentioned three fields will now be displayed at the top of the all the tabs by default:

   • First Name

   • Last Name

   • Date of Birth

   These fields will be pinned to the top regardless of whether they contain values, ensuring consistent visibility and access for all documents.

Advanced Options

The Data Review page has additional features that enable the user to search, filter, and review the data efficiently, those features are:

• Search Icon: Use the search icon to find specific values in the extracted data.

• Color-coded flags: The color-coded flags marked on the data field are based on the below-mentioned conditions:

Flags	Keys

• Low confidence Badge: The low confidence badge is used to indicate data in the Ushur system that has been predicted with low confidence and may not be accurate. The fields with low confidence badges should be reviewed and modified by humans if necessary to ensure accuracy. For more information on how to read the confidence badge, refer to How to understand Ushur Confidence Levels?.

How to review and correct the data using the Human-in-the-Loop feature?

1. Navigate to Dashboard > Data Validation tab.

2. From the Workflow drop-down menu, select the relevant workflow.

3. Once you select the workflow, the data summary table provides a detailed summary of workflow engagements.

   

4. Select the engagement where the status is marked as Requires Review.

5. On the Data Review page, click the Data in Review tab on the left panel. The data fields available under this tab need to be reviewed and edited if necessary.

6. Click the data fields to locate the source of the data from the email or PDF that is displayed in the right panel.

7. Click the field to edit the value. The Value-Modified icon will be displayed.

8. To reset the field's value to its original extracted value, click the ResetIcon displayed on the right of the field.

9. Once you complete the changes, click the Submit Changes icon. The Submit Changes icon displays the total number of changes.

Note

If no changes are required, use the More icon on the data validation summary table to complete the submission.

10. If you submit the changes before reviewing all data, a confirmation message will appear with the count of pending reviews. You can then click the Complete Review icon to finish or cancel.

    

11. If all the fields are addressed, click the Complete Review icon to proceed with the pending workflow execution.

    

Note

When making edits to data fields, ensure that the values are input in the format required by the final system processing the extracted data.

For instance, if you need to change a field value from "$7,500" to "$8,000", enter it as "8000" if the system accepts only numeric values without symbols or commas.

It's important to adhere to these format requirements, as any deviation will not be automatically adjusted and may lead to data inconsistencies.