Entity models depend on entity mentions extracted from the text through the Rosette /entities endpoint, provided by Rosette Entity Extractor (REX). REX is pre-trained to identify common entity types. Your event model may require additional entity types. In this case, you would train a custom entity model to use for event extraction.
Training Event Models with Custom Entities
To train event models using custom entity models:
With Model Training Suite, train the custom NER model. This model must be complete before you start training your event model.
Upload the custom NER model to the Rosette Server instance used by your MTS training environment. We recommend using a custom profile.
Create a new events project.
Edit the project schema. The profile id drop-down will list all custom profiles on the Rosette Server instance currently used by the MTS training environment.
The profile id must be the identifier of the custom configuration where you uploaded your custom NER model.
Annotate and train your event model.
By specifying the profile id in the project schema, you will have access to all custom entity types defined in your custom NER model.
Extracting Events with Custom Entities
To extract events that include custom entities:
Upload the trained NER model to the SAME custom profile as used for training, on the production instance of Rosette Server.
Upload the trained events model to the production instance of Rosette Server.
Call the /events endpoint.
The custom NER model must be in the same custom profile used for training.