Validate a document package

This article explains how use the Document Analyzer UI to review, validate, and, when necessary, correct the document classification and data extracted by the AI analysis.

Note

AI analysis is initiated using a series of APIs. For more information see the API reference in the Lightico developer center.

The Document Analyzer AI processes and extracts data automatically, but human validation remains an integral part of the Workflow. As the AI model is trained and refined, the need for human intervention may decrease, allowing for a more automated process. However, human review ensures oversight and provides the flexibility to validate and adjust AI-generated results when necessary.

Start the validation

Access to the Document Analyzer UI is gained through a URL provided in the url parameter of the Documents analysis started webhook event. For more information, see Document Analyzer events in Lightico developer center.

To start the validation:

• Click the link to the document package provided by your administrator.

The Document Analyzer UI opens to the document package page.

Document package page

The following information is displayed for the documents in this package:

1. ID - Unique string for this package assigned by Lightico when you initiate the analysis with the Create document analysis endpoint
   

2. Creation date - Date that the AI analysis for this package was performed
   

3. Title - Name of the document analysis type selected for this package in the Create document analysis endpoint.
   

4. Client information - Information drawn from Lightico Attributes that were defined in the Create document analysis endpoint.

   Tip

   An administrator can customize what type of information is displayed when they set up this analysis type.

5. Documents table - Information and status of the documents in this package according to category:

   • Required - Required documents according to the package rules

   • Optional - Documents that were classified, but are defined as optional for this analysis type

   • Unclassified - Documents that the AI model does not recognize and, therefore, cannot classify
     

6. Open issues - Extracted information that does not meet the criterion of the rules for this type of analysis
   
   

Document table

The document package table presents the following information for each document:

• Name - Type of document according the AI recognition

• Classification confidence - An estimate of how likely it is that the document type was identified accurately

• # of pages - Number of pages in the document file

• Last viewed - Last time it was viewed by the human analyst

• Status - see below

The following statuses appear in the document table:

• Pass - All information is complete and meets the requirements of the rules for this document

• Fail - There is missing information or some of the information does not meet the rules requirements

• Missing - (only shown for required documents) Document was not provided in the document upload for this package

• Waiting for review - Human reviewer has not completed their review and validation

• Completed - Human reviewer marked this document completed

Validate the AI analysis of each document

Validation requires a co-pilot user to approve, reject or correct different aspects of the AI analysis. There are three aspects of the AI analysis to validate for each document:

• Classification - The AI model identifies the type of document, for example, pay slip, utility bill, or bank statement. Every analysis package has defined documents types on which the AI model was trained.  

• Rules - Whether the extracted data meet the criterion of the analysis rules

• Extracted data - Individual data points extracted and identified by the AI model. Each data point is assigned to a Lightico Attribute.

Note

The following is only a suggested order of how to perform the different aspects of the validation.

To validate the AI analysis of a document:

1. Select a document in the document table.
   
   
   The document page opens.
   At the top of the right pane is the Document classification given to this document and a confidence score.

2. If the classification is correct, click the thumbs up.

3. If there are multiple pages of this document, do this for all the pages.
   
   

4. Click the Rules tab in the right pane.
   If the document has a status of Fail, the rules pane shows the reason for the failure.

   
   

   Note

   Validation is for verifying that all classification and extracted data is accurate. A document that failed to comply with one of the analysis rules can still be validated and submitted. The submitted analysis report will contain the information that this document did not comply with the rule.

   
   

5. Select a piece of data.
   The data is highlighted in the document view.
   

6. Review each piece of data and either:

   1. Click the thumbs down to reject and correct the extracted information (see below).

   2. Click the thumbs up to approve the information.  
      

   
   The Submit button at the top of the page becomes active when all data has been validated by clicking one of the thumb buttons.

7. Click Submit.
   
   
   The document is locked and marked as completed. No further validation or editing can be done.

Rejecting and correcting extracted information

If you see that a piece of data was extracted incorrectly, you can reject it and change it to the correct information.

1. Click the thumbs down icon.
   
   

2. Enter the correct version of the data in the box and click Done.
   
   

   The information is recorded in this document as you corrected it and it is marked as Edited.
   

Completing the package validation

Once you have classified, validated, and summitted all documents, the package validation is complete.

Submitting unclassified documents

The package may contain documents that the AI model could not associate with any of the known document types in this package.

To complete the package validation click into each of the unknown documents and either give select a document type or submit the document as unclassified.

Obtaining information about completed analysis and validation

Lightico provides information about completed document packages through the following webhook events.

• For full documentation of these events, see Document analyzer events in the Lightico developer center.

• For information about defining webhook event destinations, see Custom integration hub for API and events.

Document validation completed event

This event is sent when a user submits a document after manually reviewing and validating it in the Document Analyzer UI.

It provides data about the classification, extracted information and any manual changes that were made.

Document analysis summary event

This event is sent when all the documents in the document analysis package have been reviewed and submitted by the user in the Document Analyzer UI.

It identifies any documents that are missing from the package according to the analysis rules.