Results for fine-grained entity typing. All LITE model results are statistically significant (p-value ¡ 0.05 in t-test) in comparison with the best baseline results by MLMET on OntoNotes and by SEPREM on FIGER.
| Model . | OntoNotes . | FIGER . | |||
|---|---|---|---|---|---|
| macro-F1 . | micro-F1 . | macro-F1 . | micro-F1 . | ||
| Hierarchy-Typing (Chen et al., 2020b) | 73.0 | 68.1 | 83.0 | 79.8 | |
| Box4Types (Onoe and Durrett, 2020) | 77.3 | 70.9 | 79.4 | 75.0 | |
| DSAM (Hu et al., 2020) | 83.1 | 78.2 | 83.3 | 81.5 | |
| SEPREM (Xu et al., 2021) | – | – | 86.1 | 82.1 | |
| MLMET (Dai et al., 2021) | 85.4 | 80.4 | – | – | |
| LITE | pre-trained on NLI+UFET | 86.6 | 81.4 | 80.1 | 74.7 |
| NLI+task-specific training | 86.4 | 80.9 | 86.7 | 83.3 | |
| Model . | OntoNotes . | FIGER . | |||
|---|---|---|---|---|---|
| macro-F1 . | micro-F1 . | macro-F1 . | micro-F1 . | ||
| Hierarchy-Typing (Chen et al., 2020b) | 73.0 | 68.1 | 83.0 | 79.8 | |
| Box4Types (Onoe and Durrett, 2020) | 77.3 | 70.9 | 79.4 | 75.0 | |
| DSAM (Hu et al., 2020) | 83.1 | 78.2 | 83.3 | 81.5 | |
| SEPREM (Xu et al., 2021) | – | – | 86.1 | 82.1 | |
| MLMET (Dai et al., 2021) | 85.4 | 80.4 | – | – | |
| LITE | pre-trained on NLI+UFET | 86.6 | 81.4 | 80.1 | 74.7 |
| NLI+task-specific training | 86.4 | 80.9 | 86.7 | 83.3 | |