Table 5.

Results using different neural networks and document presentations detecting suspicious cases of COVID-19.

Model	Doc. Repr.	SNOMED-CT Repr.	P (%)	R (%)	F1 (%)	Acc.(%)	AUC (%)	MCC (%)
Basic ANN	TF-IDF n-grams	Baseline	79.57	81.05	79.55	81.49	81.05	63.23
	disable IDF	TF-bin + MI	85.66	85.23	85.28	85.54	85.24	70.90
CNN	FastText SUC	Baseline	84.17	84.52	83.66	84.03	84.52	68.67
		TF-bin + MI	85.17	84.96	84.47	84.75	84.96	70.11
LSTM	FastText SUC	Baseline	80.22	77.45	75.77	76.59	77.45	57.46
		TF-bin + MI	82.66	81.01	79.11	79.69	81.01	63.51
BiLSTM	FastText SUC	Baseline	78.01	74.40	71.63	73.00	74.40	52.01
		TF-bin + MI	78.92	76.45	75.41	76.62	76.45	55.09