Table 7 Doc-level performance on the CDR dataset with different scales of CTD knowledge
From: Biomedical relation extraction via knowledge-enhanced reading comprehension
KB(CTD)\(\_\)Ratio | 0.25 | 0.50 | 0.75 | 1.0 |
Document-level F1 (%) | 63.68 | 64.81 | 67.09 | 71.18 |