We are working on an NLP use case where entity extraction from emails with predefined training data is working well. Now, we want to allow end users to contribute to training the NLP model.

In Pega Customer Service, users can right-click on text within an Email Case to select or correct entities. When this is done, training records are automatically generated in the Email channel’s Training Data tab.

However, this feature does not occur in a standalone Email Bot application built with Theme Cosmos.

In Customer Service, this functionality is handled by the pega_cpm_IP_EventManager JavaScript file, which manages entity mapping updates and triggers the creation of training data records in the Email channel.

How can we implement similar training data generation in a non–Customer Service Email channel setup? Any suggestions or best practices would be greatly appreciated.