Table 4. Classification results for fine-tuning BERT models on the full dataset.
CLS means classification task.
| Model | Task | Overall Acc. |
Positive precision |
Positive recall |
Positive F1-score |
|---|---|---|---|---|---|
| BERT | Rej. CLS | 0.945 | 0.000 | 0.000 | 0.000 |
| BioBERT | Rej. CLS | 0.953 | 0.697 | 0.447 | 0.515 |
| Clinical BERT | Rej. CLS | 0.977 | 0.923 | 0.632 | 0.750 |
| Kidney BERT | Rej. CLS | 0.977 | 0.867 | 0.684 | 0.765 |
| exKidneyBERT | Rej. CLS | 0.978 | 0.811 | 0.789 | 0.800 |
| Model | Task | Overall Acc. |
Weighted precision |
Weighted recall |
Weighted F1-score |
| BERT | IFTA CLS | 0.702 | 0.690 | 0.697 | 0.679 |
| BioBERT | IFTA CLS | 0.731 | 0.729 | 0.731 | 0.726 |
| Clinical BERT | IFTA CLS | 0.733 | 0.734 | 0.732 | 0.729 |
| Kidney BERT | IFTA CLS | 0.730 | 0.731 | 0.722 | 0.724 |
| exKidneyBERT | IFTA CLS | 0.714 | 0.710 | 0.712 | 0.708 |