Enable a pretrained FERN model

Learn how to enable a pretrained FERN model to evaluate matches between pairs of entities

Prerequisites:

You must have ROLE_ADMIN_TENANT Access permissions
You must have a custom role with MDM:config.businessModel UPDATE permission and CatalogService with READ permission.
Note: If the custom role assigned to the user does not include CatalogService with READ permission, the system cannot retrieve the pretrained model information required to configure FERN modes.
We've created ready-to-use pretrained FERN models for certain entity types/industries. To use them, you need to fulfill the specific set of conditions for the pretrained FERN model you want to enable. For details, see topics Available pretrained FERN models and Map attributes for pretrained FERN models.

Use a pretrained FERN model to evaluate matches between pairs of entities. For more information, see topic Use a pretrained FERN model.

To enable a pretrained FERN model:

In the Reltio app selector, navigate to and select Console > Data Modeler > Entity types.
Select the entity type for which you want to set up the pretrained FERN model.
Go to the Match tab.
If there are pretrained models available for the entity type you've selected, they show up in this area.
In the AI-Powered Models section, click anywhere on the row of the entry to view its details. Click the pencil icon to edit details of the entity type.
Enable the AI-Powered model option for the individual-like objects.
By default, the Status option is displayed as "Enabled". Choose Internal and/or External on which you want to perform match actions. Select Match by Operational Value (OV) only if you want to perform match action by OV.

Set relevance match score – Set match action and define a corresponding Relevance score range:


Match action	Defines the match action – "Automatically merge" and "Potential matches". Note: Select "Automatically merge" option if you want to merge the record automatically if it matches with the relevance score range as defined. Select "Potential matches" option if you want to manually confirm from the possible matches from the record.
Action label	Defines a label for the match action that you selected.
Relevance score range	Defines the relevance score range for the match action that you selected. Note: The relevance score must overlap from first match action to the second and the second match action to the third match action.

By default, you can set three match actions. The relevance score range indicates the lowest and highest values that result in a match. The highest Relevance score range is defined in the first Match action. The Relevance score range in the second Match action must match with the lower limits in the first Match action. The Relevance score range in the next Match action must match with the lower limits of the second Match action.

For example, if you define a range of 0.9 to 1 for the first Match action, in the second Match action your higher value must be 0.9. If you define a range of 0.75 to 0.9 for the second Match action, the higher value must be 0.75 in the next Match action.

Select Map required attributes:

Note: This section shows as expanded if it's selected as "Enabled" in the first drop-down field. If you don't see the mapping attributes details, click the this section title. You can map your entity type attributes to the corresponding FERN model attributes.
1. Map the attributes in your data model to the corresponding Required attributes (mandatory) FERN model.
  Select an attribute drop-down from your entity type attribute and select multiple attributes that you want to map to the Required attribute. For the Individual entity type, the following screen opens: For the Organization entity type, the following screen opens:
2. Optionally, map the attributes in your data model to the corresponding Other attributes FERN model attributes.
Save your changes.
Run the Rebuild Match Table task.

The pretrained model is now enabled and will start evaluating match pairs.

Unify and manage your data

Enable a pretrained FERN model

Footer