search

mandarjoshi90 / coref

BERT for Coreference Resolution
Apache License 2.0
441 stars 92 forks source link

Bert Retraining #49

Closed bodasheera closed 4 years ago

bodasheera commented 4 years ago

Hi Mandar ,

While running spanbert large model , if is not able to capture some of words such as Degree . The same text if I use other brand like Kurkure it will detect . Can I train coref model with my own dataset .

How do I do it ? I have access to onto notes 5. I'm new in the field on AI . Should I convert my dataset to ontonotes format? How can I train on custom dataset.

fairy-of-9 commented 4 years ago

Convert your dataset to conll format. (I think It's easiest way) and convert conll to jsonlines using minimize.py you can see conll format through results of ./setup_training.sh