Skip to main content
. 2020 Jul 31;8(7):e17638. doi: 10.2196/17638

Table 5.

An ablation study for our model.a

Model Intrasentence level Intersentence level Relation merging
Precision (%) Recall (%) F-measure (%) Precision (%) Recall (%) F-measure (%) Precision (%) Recall (%) F-measure (%)
Without multihead attention 58.2 82.9 68.4 44.7 44.3 44.5 55.1 71.3 62.2
Without GCNb 62.6 74.1 67.9 43.6 48.4 45.9 57.1 66.4 61.4
Our model 59.1 81.5 68.5 47.8 52.2 49.9 56.3 72.7 63.5

aThe values in italics indicate significant findings.

bGCN: graph convolutional network.