Hello @asynadinou,
This is a great question, and a problem I've had historically. I trained several different entity extraction models, but kept running into problems with accuracy. Have you tried creating your own AI Prompt in AI Builder, and are you able to? Since moving my entity extraction models to custom prompts, the accuracy of finding different variations is much higher. I've assume this is due to the underlying language model running the entity extraction model versus the GPT model.
I hope this help! If it did, please feel free to accept my response as the solution 🙂
-Reid
