Hey there!
I have a pipeline with some custom entities. I've added examples for the NER pipe, and added an EntityRuler. In the end I have a TextCategorizer.
I'd like to have the text categorizer use the output entities found by the NER and EntityRuler, and it seems that the regular training of the models doesn't do that. What is the best way to achieve that?
The TextCategorizer doesn't use the NER annotations as features. You could write a custom model that did that if you want. There's no single obvious way to use the annotations though --- it would depend on what you want the model to learn, and you'd have to experiment.
Thanks for the reply. What I'm looking to create is something similar to an intent predictor where I can use the output of the slot-filling (NER-based) for the intent classification.