Skip to main content
Our proprietary sentiment scoring, entity resolution, and speaker identification models each undergo continuous validation to ensure accuracy, reliability, and monitor model ‘drift’. If over time considerable language drift is detected, such as by the emergence of new topics or terms (e.g., “COVID”), a scoring update may be triggered. When an update occurs, we take the following steps:
  1. Increment the model version
  2. Update any discrepant historical rows
  3. Announce the end-of-life for the previous version in release notes
  4. Maintain dual models during the transition for seamless integration
This process ensures that our clients always have access to the most accurate and up-to-date data while maintaining backward compatibility during transition periods.

Need Help?

Have questions about a specific update or need assistance with model transitions? Contact our support team for detailed guidance.
⌘I