To add another option - use Spacy’s pretrained PERSON entity, either on it’s own or in conjunction with your own trained name entities. This will help pick up many more names than you include in your NLU data.
yes, you are right, if I am not wrong it happens because the Spacy library is already trained with many names.
but the library is developed with american and british names, whereas my use case i mainly for Indian names. extracting those names spacy doesnt help properly.
btw i have a new use case to train an individual end to end NER model specifically extracting Indian names, for training purpose i have 15k Indian names but dont know how i should train on them in order to extract the Indian names from any sentence.
Gotcha, that makes sense! You can take a look at spacy’s documentation for training your own models here: https://spacy.io/usage/training
If you do that, please post and let us know how it goes!