Table 5.
Performance of the deep joint learning method on each type of FM information and relation
| Type | P | R | F | |
|---|---|---|---|---|
| FM information recognition | FM (Maternal) | 0.9412 | 0.9552 | 0.9481 |
| FM (Paternal) | 0.9286 | 0.7800 | 0.8478 | |
| FM (NA) | 0.8452 | 0.8875 | 0.8659 | |
| Observation | 0.8753 | 0.9146 | 0.8945 | |
| LSa | 0.8418 | 0.9116 | 0.8753 | |
| Overall | 0.8775 | 0.9030 | 0.8901 | |
| Relation Extraction | FM-LS | 0.6084 | 0.6273 | 0.6177 |
| FM- Observation | 0.6451 | 0.6451 | 0.6451 | |
| Overall | 0.6327 | 0.6392 | 0.6359 |
aThe results are obtained according to the gold LS mentions, not the gold standard LSs for final evaluation, which are not provided. Therefore, the overall performance on FM information recognition does not cover LS