Quality difference in NER between dicta-il/dictabert-large-ner and dicta-il/dictabert-joint

#1
by MaayanHive - opened

Hi,

Will using this model yield better results over the joint model?
We wonder if we should consider using only the lex of dictabert-joint and discard its NER.

Thanks

DICTA: The Israel Center for Text Analysis org

Hi Maayan,

You can read more about the model and it's scores here: https://arxiv.org/abs/2403.06970

The dictabert-large models give a performance boost and yield better results than the dictabert-base models. This dedicated NER model performs marginally better than the joint model - while the joint model has the advantage of performing all the tagging in one go.

Sign up or log in to comment