Veritas Surveillance Reviewer's Guide

Viewing Intelligent Review Details

The Intelligent Review Details section provides the facts of why the item is classified as Unreviewed Relevant or Unreviewed Irrelevant. It shows the Relevant and Irrelevant labels (links) and the respective contribution.

During review, the Intelligent Review Details section appears on the Preview tab only if the departments you want to review are enabled for Intelligent Review, and the Show Intelligent Review Details in Review permission is enabled for the logged-in user. This permission is by default enabled for the Department Reviewer, Escalation Reviewer, Compliance Supervisor, Exception Reviewer, and Passive Reviewer roles, where either the Review Messages or the Review Escalations permissions are enabled. By default, Intelligent Review Details section is collapsed. Users can expand and collapse it as required.

The total relevant and irrelevant contribution value is shown besides the respective labels. These values (between 0 to 100) are factor of relevant and irrelevant contributions found inside the item. When you click the Relevant and Irrelevant labels, the corresponding details appear which shows the factors that have contributed towards relevant or irrelevant. The calculated values of a contribution of each factor are mentioned so that a reviewer can understand the reason behind the item being relevant or irrelevant.

If the contribution value of the relevant factors is more than the contribution value of the irrelevant factors, the item will have higher relevance score (greater than 50). If the contribution value of the irrelevant factors is more than the contribution value of the relevant factors, the item will have lower relevance score (less than 50). If the contribution values of the relevant and irrelevant factors are almost similar, the item will have the relevance score around 50.

The factors that contribute are Author, Recipient, Subject, Content, Direction, Tags, and Department influence. Department Influence shows the extent to which prediction is inclined in favor of or against the marking based on the department's learning. If reviewers in a department favor marking items as irrelevant, then Department Influence will contribute towards irrelevance and vice versa. If factors do not have values, they can still contribute as relevant or irrelevant. It happens when the algorithm weighs the lack of details (such as the presence of zero tags or absence of content) as a contributing factor to its learning.

If the contribution of any factor is insignificant, but not absolutely zero (0), then value can be rounded to and displayed as zero (0). If the factor does not contribute at all, then this factor is not displayed. The amount of contribution of each factor is also shown with the color legend.

Refer to the sample images below.

Figure: Preview pane showing the IR details, hit highlighting, managing text for learning, and AI-predicted labels

Figure: Contributing factors and their contribution

Following are some circumstances when the Intelligent Review Details are not available, and the application displays different messages for users.