Changed behavior

You can now remove the match fields that were added in the following scenarios from the Adaptive AI match model training process:

Previously, you couldn’t remove these fields from the Adaptive AI match model training process.

For more information about training an Adaptive AI match model, see Training Adaptive AI match models.

When you train Adaptive AI match models, you can no longer label record pairs as Need More Data.

Previously, the Need More Data label was available for labeling during match model training.

For more information about training an Adaptive AI match model, see Training Adaptive AI match models.

You can now control which user roles to train an Adaptive AI match model. By default, no user roles are allowed to train models.

To allow a user role to train models, assign the Adaptive AI Match Model privilege to the user role on the Security page.

If you have users who previously had access to train these models, ensure that you update their custom user roles with the new privilege.

Previously, the following predefined user roles had the privilege to train Adaptive AI match models:

For more information about configuring privileges for training an Adaptive AI match model, see Configuring privileges for training Adaptive AI match models.

When you train an Adaptive AI match model, the data access rules configured for a business entity control your access to the training data set associated with the model.

Previously, you could access the training data set associated with an Adaptive AI match model irrespective of the data access rules configured for the business entity.

For more information about the impact of data access rules on Adaptive AI match models, see Data access rules on Adaptive AI match models.

The uniqueness criteria property of field groups has been removed from the general properties.

Previously, you could view uniqueness criteria as a general property of field groups, along with a message indicating that it has been moved to the Survivorship page and renamed Deduplication.

For more information about deduplication criteria, see Deduplication criteria.

To improve performance, the match process now groups record pairs that match indirectly without any explicit rules. For example, if record A matches B and record B matches C, the match process creates record pairs AB, BC, and AC. The record pair AC matches indirectly without any explicit rules. The match process then groups these record pairs into a single group containing records A, B, and C.

Previously, the merge process handled the grouping of record pairs that matched indirectly into a single group.

The responses of the Read Master Record by Business ID and Read Master Record by SourcePKey REST APIs don’t include the survival parameter in the ruleAssociation array.

Previously, the ruleAssociation array included the survival parameter.

For more information about the Read Master Record by Business ID and Read Master Record by SourcePKey APIs, see Business entity record APIs.