Model languages and number of Training records
Somewhere I read that this is best model for French language, does this works well with English ?
In order to train this model with custom entities, how many training records needed per Entity ?
Hello Ashish,
This is based on Camembert model which is a language model trained specifically for french. So performance for English will probably not be good.
For English you can use model like Bert or Roberta, and train them on your entities. It's difficult to give you a number of records required per entities, sometimes a few hundreds are enough to have correct results.
Also you can have a look at model already trained for ner for English, for instance I trained this one if you want to try it:
Jean-Baptiste/roberta-large-ner-english
But there is plenty of other also available on HF for English.
Thanks,
Jean-Baptiste
Thanks a lot Jean, this info is very helpful!
I will try the model you suggested.
Thanks,
Ashish