Labeling / Training

---

Item Labeling

Quickly click findings to trigger retraining.

Action labeling options

The following action labels instruct Collibra on how to handle a finding:

Action	Description
Validate	Instructs Collibra to either assign a finding to a specific user for review, which then appears in the View the Assignment Queue, or acknowledge without an assignee that the finding is a valid observation. Note: Validating a finding does not improve your score.
Invalidate	Instructs Collibra to ignore a finding and allow the value to pass. There are two invalidation options: Save and Save & Retrain. Save: Allows you to mark a finding as invalidated. Save & Retrain: Allows you to invalidate a finding and any previously saved invalidated findings (if any). Note: When you have many findings to invalidate, it may be best to use the Save option to invalidate them at the same time, once all findings are reviewed.
Resolve	Instructs Collibra to mark the finding as an observation and prevents it from appearing in future runs. Resolving a finding does not immediately affect data quality scores.

Available actions by feature

Feature	Available actions
Behaviors	Validate, Resolve
Rules	Validate, Resolve
Outliers	Validate, Invalidate, Resolve
Pattern	Validate, Invalidate, Resolve
Source	Validate, Invalidate, Resolve
Record	Validate, Resolve
Dupes	Validate, Invalidate, Resolve

Warning Some findings are ineligible for all labeling options. For example, you can only apply Validate and Resolve labels to findings that result from Rules.

Validating a finding

When you apply a validate label to a finding, you can assign it to another DQ user to review. This marks the finding for future runs and sends it to the internal Assignments Queue. You can also configure DQ to send assignments to an external queue, such as the ServiceNow tool. For more information, go to Manage Collibra DQ Assignments.

Invalidating a finding

Sometimes the findings page flags issues with your data that DQ discovers during a job run, but maybe you want DQ to ignore certain flagged issues. The invalidate label allows you to do that. After you add a descriptive annotation of your action, you can then select either Save or Save & Retrain.

Save

If you have a large number of findings that DQ has flagged, and you want to invalidate all of them at once instead of clicking through one at a time, select Save for all of the findings you would like to bulk invalidate. On your last finding, select Save & Retrain. All previously saved invalidated findings are removed and DQ retrains your dataset.

Save & Retrain

When you Save & Retrain your dataset, any previously deducted points from a flagged finding are restored and reflected in your overall data quality score. If you do not have many findings to invalidate, you can Save & Retrain individually instead of in bulk.

Resolving a finding

Some features, such as Behavior and Rules, only permit Validate and Resolve actions. When you cannot Validate a finding but you want to apply a label, select Resolve. The Resolve label prevents a finding from appearing in future runs of your dataset, and does not immediately affect your data quality score when applied.

Recalling labeled findings

To modify a previously labeled finding, you can always access them through the Labels tab. Here you can edit an annotation or delete a label entirely. If you delete a label, it returns to the findings page, unlabeled. From there you can again choose to Validate, Invalidate, or Resolve it.

To closely analyze when a finding has received a label, who has applied it, and more, see also the Dataset Audit Trail.