what is the maximum limit of entities we can have in a spacy or bert based custom NER models ? I have seen examples over the web which have been trained to a max of 10 custom entities per model and even the Ontonotes data set has 18 entities, so does having certain number of entities in a NER model can impact the model learning abilities and performance ?
Asked
Active
Viewed 204 times
1 Answers
2
I don't think there is a certain limit but increasing of number of classes would definitely effect the learning in a negative way. You can train a model with a 50 classes but the learning performance is depends on how your your dataset represent each class (is there a class imbalance problem etc. ) and whether your classes are distinguishable from each other easily or not.

Inputvector
- 1,061
- 10
- 22
-
2For the record, there is no hard limit in spaCy. I would be surprised if other libraries had a hard limit either. – polm23 May 09 '22 at 01:52