training model with own data about spanbert HOT 3 CLOSED

Not sure I understand. If this is about relation extraction, we used the TACRED dataset.
https://nlp.stanford.edu/projects/tacred/

from spanbert.

Hi @mandarjoshi90 ,
Thanks for your reply. I understand that you have used TRACRED, but my question was about how you formatted that data. I mean how do you represent entities and data together to train your model.
for eg: surrounding the entities with an special entity specific embedding or something else?

from spanbert.

@platinum736 Please see the implementation here: https://github.com/facebookresearch/SpanBERT/blob/master/code/run_tacred.py#L116-L221

We implemented four different feature modes: text, ner, ner_text, text_ner.
And we only used "ner" mode at the end for the TACRED results (use NER tags to replace entity text).

Hope this helps.

from spanbert.