Table 5.
F1 scores of coreference on CRAFT test set in comparison with some baselines
| System | B3 | BLANC | CEAFE | CEAFM | LEA | MUC | Ave |
|---|---|---|---|---|---|---|---|
| E2E_MetaMap | 36.4 | 46.5 | 33.1 | 41.0 | 32.4 | 51.8 | 40.2 |
| BERTfilter | 44.0 | 48.9 | 39.8 | 49.0 | 40.0 | 57.0 | 46.4 |
| BioNeu | 45.0 | 55.4 | 36.1 | 49.8 | 41.8 | 55.1 | 47.2 |
| BioNeu-feature | 45.3 | 53.2 | 36.5 | 49.4 | 42.1 | 56.1 | 47.1 |
| BioNeu + SFA | 45.1 | 56.2 | 37.0 | 49.7 | 42.0 | 56.3 | 47.7 |
| KE-LSTM | 54.9 | 63.1 | 48.6 | 59.4 | 51.3 | 64.5 | 57.0 |
| KE-LSTM-feature | 54.5 | 62.2 | 48.1 | 59.2 | 51.4 | 64.5 | 56.6 |
| KE-LSTM + SFA | 55.0 | 63.6 | 49.5 | 59.4 | 51.7 | 64.6 | 57.3 |
E2E_MetaMap and BERTfilter are the baselines in [7]
The maximum value is in bold