Model evaluation
miladnouriezade opened this issue · comments
Hi, I've tried to evaluate your model with scibert pico data set, but the tokenization wasn't the same, could you please share the F1 score you get?
Hi @miladnouriezade. Thank you for your interests. The performance of a part of the entity classes including PICO is listed below (note that we have a new set of attribute categories for PICO and delete the original "modifier" and "measure". we will release our new version in the future):
Intervention: precision: 69.14%; recall: 73.68%; FB1: 71.34
Observation: precision: 70.54%; recall: 80.53%; FB1: 75.21
Outcome: precision: 71.51%; recall: 78.85%; FB1: 75.00
Participant: precision: 58.11%; recall: 70.49%; FB1: 63.70