Table 3.
Performance on test set for relation (Track 2) and end-to-end (Track 3) extraction task of submitted and improved models.
| Model | Precision | Recall | F1-score |
|---|---|---|---|
|
Relation |
|||
| * Intra [ensemble] + Inter [ensemble] | 0.9463 | 0.9480 | 0.9472 |
| Intra [ensemble] + Inter [ensemble] | 0.9572 | 0.9456 | 0.9514 |
|
End-to-end |
|||
| * NER [recall] + Weighted [ensemble] + Inter [ensemble] | 0.9264 | 0.8318 | 0.8765 |
| NER [recall] + Intra [ensemble] + Inter [ensemble] | 0.9286 | 0.8321 | 0.8777 |
The asterisk indicates our submitted models to the n2c2 shared task.
NER: named entity recognition.