Entity Classification Migration Guide

The Entity Classification Service (V2) experience introduced a reworked classification hierarchy, seeded detector types, and simplified likelihood buckets so users can achieve higher accuracy, better address common use cases through flexible configurations, and readily understand the classification results presented in the application with more intuitive UI.

The goal of the V1 to V2 migration effort is to minimize disruption of active classifiers, classification profiles, data source settings, and classification setting preferences. Listed below are the expected changes with V2 in more details.

The V2 engine was announced as the default classification experience for new tenants in the OneTrust Spring 2024 Release (202405.2.0). This requires no preview feature enablement.

For existing clients on the V1 engine prior to this release, please contact OneTrust Support to request the migration of your account to the V2 engine.

All of the classifiers present in V1 are available in V2. Most of those classifiers have the same name. However, there are a few classifiers in V1 which were combined into 1 single classifier in V2. Additionally, a classifier in V1 may have multiple classifiers in V2. This was done to increase recall and precision of the recommendations. The impact of this is minimal on the user as the logic is handled by the system during migration. More details are listed in each section.

New system classification profiles have been updated to more accurately reflect the use case of the profile. These updates are made to reflect the broader coverage with regard to entities that can be classified using V2 as well as higher accuracy of the system classifiers. The changes include adding new classifiers which were available only in V2, reducing thresholds etc.

In addition, thresholds and samples now exist on the classification profile in V2 (unlike at classifier level in V1). This will also be honored for profiles which inherit those values from the threshold in the classifier(s) in V1.

All of the migrated system classification profiles will reflect the improved classifiers' updates.

• If a classifier in V1 (say US Address) was improved for higher accuracy in V2, the profile will reflect the updated classifier with higher accuracy after the migration.

• If there a mismatch between system classifiers in V1 vs V2 classification profile, e.g. V2 classification profile has an additional classifier which is newly added and only available in V2 classification engine, the system classification profile will reflect the change after the migration

• If a V1 system classification profile had a classifier which was merged or split, the V2 system classifier will consume the classifier which the V1 classifier was merged or split into.

• Any classifiers that were added to a V1 system classification profile will be overridden with the migration. The V2 classification engine does not support edits to system classification profiles, hence the updated system profile needs to be copied and modified as a custom classification profile, then added to the data source in V2.

All classification profiles are automatically linked to the corresponding data sources. Term results from previously scanned data sources continue to be available after migration.

• System and custom classifiers will continue to be indexed based on the V1 setting of the classifier.

• If a classifier is split or merged, the results will be indexed per new individual classifier, and the label annotation will maintain the same naming convention to enable search and retrieve capabilities. For example, if a Health classifier was indexed in V1, all of the classifiers (drug name, disease name etc.) would be indexed in V2.

The Allowlist is designed to honor the term and data source, not a specific classifier. The same logic is consistent with the V2 engine. However, a refreshed auto-approval settings page is on the roadmap which would enhance the allowlist functionality.

All changes to classifiers will continue to map onto the originally mapped term. There are no changes to the term entity as part of the V1 to V2 migration.

The push of data to Data Mapping Automation as data elements occurs through the term-to-data element mapping. The change in classifiers should not change this process as long as the term mapping remains consistent.

The new Quick Review screen will show a table which includes a column for the classifier used for discovery so users can better identify the respective term that is also available for review.

The reviewed recommendations from existing scans will continue to be available. However, the confidence scores would be converted to likelihood in the results. Confidence scores of system classifiers will change to likelihood. This is done so that the recommendations have a specific action item based on the likelihood.

For example, if a recommendation is “Very Likely”, this recommendation can be auto-approved with minimal review. Confidence scores for custom classifiers would map to the likelihoods as per the table in the section below.

A new review screen to leverage the V2 engine capability in order to optimize the review process is currently available as well as a different roadmap feature. The new review screen features a Quick Review screen that would show samples aggregated across all data assets in a single view with an option to view context surrounding the unstructured data. 

For additional information or if you have questions about this migration, please contact OneTrust Support.