Table 3 Test results of relation extraction task on benchmark datasets using micro-averaged F1-score
Method/Dataset | ADE | BioInfer | i2b2-clinical | i2b2-2012 | CHEMPROT | JNLPBA | N2C2 |
|---|---|---|---|---|---|---|---|
C4.5 DT [35] | 71.30 | 64.17 | 69.61 | 69.32 | 77.24 | 67.97 | 65.59 |
BiLSTM-CRF [36] | 80.13 | 82.10 | 76.93 | 73.60 | 78.14 | 76.09 | 71.35 |
BiLSTM-CNN [78] | 79.74 | 74.40 | 68.91 | 72.13 | 73.44 | 72.41 | 70.10 |
RNNs [36] | 80.10 | 82.20 | 80.15 | 78.04 | 77.00 | 78.14 | 75.57 |
CMAN [37] | 81.10 | 73.10 | 68.44 | 74.32 | 73.70 | 66.65 | 68.50 |
Adversial [37] | 75.50 | 79.60 | 69.19 | 78.10 | 72.94 | 69.34 | 69.57 |
Multi-att-CNN [79] | 79.24 | 84.35 | 70.43 | 78.80 | 77.50 | 71.62 | 70.15 |
BLUE BERT [77] | 80.39 | 82.09 | 72.32 | 79.10 | 79.98 | 80.56 | 79.10 |
BioBERT [13] | 82.03 | 86.90 | 81.20 | 80.23 | 81.46 | 83.41 | 79.18 |
Our approach | 90.00 | 88.88 | 90.12 | 85.03 | 88.95 | 84.50 | 84.10 |