Table 5.

Comparison with the baseline method.

Model			Precision (%)		Recall (%)		F-score (%)
Single models
	CNNa	42.47		69.43		52.70
	GCNb	48.77		63.69		55.24
	Bi-LSTMc	60.59		60.34		60.46
Ensemble models
	Bi-LSTM+CNN	57.77		64.73		61.05
	Bi-LSTM+GCN (our model)	63.79		66.62		65.17

^aCNN: convolutional neural network.

^bGCN: graph convolutional network.

^cBi-LSTM: bi-directional long short-term memory.