Train ner component on subset of labels in training corpus

[kinghuang]

I have a corpus that is fully labelled with the target entities. But, some of these are meant for entity_ruler components to label, while others are for an ner component to learn and predict. Is there a way to configure a ner component to only learn specific labels? Or, do I have to produce a filtered version of the corpus for it?

I see that the ner component can take a labels argument during initialization, and spacy init labels can generate the content for this. But, it doesn't look like that's an easy way for me to include or exclude certain labels.

[polm]

No, we don't have a feature for this. I think the code to create a filtered DocBin should be minimal.

[kinghuang]

Ok, I'll filter the data that goes into making the DocBin files in my project. Thanks!