Table 2 Open in new tab Comparison of... | Oxford Academic

Table 2

Open in new tab

Comparison of model effect of different network layers

Layer	S4	BioRED
AVG	94.79	95.48
MA + 1D-CNN	94.71	95.67
GCN + AVG	95.03	95.60
GCN + MA + 1D-CNN	94.65	95.69
GAT+AVG	95.02	95.26
GAT + MA + 1D-CNN(ours)	96.47	96.03

Layer	S4	BioRED
AVG	94.79	95.48
MA + 1D-CNN	94.71	95.67
GCN + AVG	95.03	95.60
GCN + MA + 1D-CNN	94.65	95.69
GAT+AVG	95.02	95.26
GAT + MA + 1D-CNN(ours)	96.47	96.03

AVG: entity average embedding method; MA: multi-head self-attention mechanism

Table 2

Open in new tab

Comparison of model effect of different network layers

Layer	S4	BioRED
AVG	94.79	95.48
MA + 1D-CNN	94.71	95.67
GCN + AVG	95.03	95.60
GCN + MA + 1D-CNN	94.65	95.69
GAT+AVG	95.02	95.26
GAT + MA + 1D-CNN(ours)	96.47	96.03

Layer	S4	BioRED
AVG	94.79	95.48
MA + 1D-CNN	94.71	95.67
GCN + AVG	95.03	95.60
GCN + MA + 1D-CNN	94.65	95.69
GAT+AVG	95.02	95.26
GAT + MA + 1D-CNN(ours)	96.47	96.03

AVG: entity average embedding method; MA: multi-head self-attention mechanism