search

stanford-crfm / BioMedLM

606 stars 64 forks source link

BioMedLm for NER and sentiment analysis #10

Open sepehrasgarian opened 1 year ago

sepehrasgarian commented 1 year ago

Hi thanks for your great work, I was wondering if the model is also fine tuned for the NER or sentiment analysis task

J38 commented 1 year ago

We haven't done this at this time, focusing more on tasks like Q+A? Is there a particular sentiment analysis task you'd like us to look into? We have a set up for sequence classification so it'd be pretty straightforward to try to apply this to sentiment.

rthapa84 commented 1 year ago

Had a similar questions. How will I go about fine-tuning on NER tasks, such as n2c2? Any suggestions/pointers are appreciated. @J38

J38 commented 1 year ago

The sequence classification code is good for sentiment, but you will need to produce training data. We haven't really focused on classic NLP tasks like NER, but there was some code for that, but I didn't refine it enough to get any interesting NER results yet.