Dataset used and model performance
caolingyu opened this issue · comments
Hi Tian,
I'm interested in your work and have some questions.
- Did you use PubMed abstracts to train the model? if so, how large is it?
- Could you provide performance of model on your test data, such as f1 score?
Thanks a lot in advance.
Hi conglingyu,
Thanks for your interests. We pretrained our model on a large dataset of PubMed abstracts and then fine tuned with our own annotation with PICO. Because this is still a nundergoing project, we will release the results once is finished.
Tian