Table 2.
Comparison of RENET2’s performance on the full-text dataset when trained with different settings
| Ensemble | SeFi | SiDa | Precision | Recall | F1 score | F1 score increment |
|---|---|---|---|---|---|---|
| N | N | N | 0.5417 | 0.6887 | 0.6064 | - |
| Y | N | N | 0.5791 | 0.7327 | 0.6469 | 4.05% |
| N | Y | N | 0.6465 | 0.7233 | 0.6826 | 7.62% |
| N | N | Y | 0.5885 | 0.6562 | 0.6205 | 1.41% |
| Y | N | Y | 0.6079 | 0.6946 | 0.6484 | 4.19% |
| N | Y | Y | 0.6888 | 0.6906 | 0.6897 | 8.32% |
| Y | Y | N | 0.6746 | 0.7606 | 0.7150 | 10.86% |
| Y | Y | Y | 0.7062 | 0.7371 | 0.7213 | 11.49% |
The best result in each column is in bold. ENS: Ensemble, SeFi: Section Filtering, SiDa: Silver Dataset, F1 increment: increase in F1 score compared to the basic setting.