Table 6.

Results of two-class assertion classification. The best scores are bolded.

Dataset	Model	Present	Absent	micro F-1
i2b2 2010	Logistic Regression	0.926	0.866	0.911
	NegEx [5]	0.925	0.836	0.906
	RadText [35]	0.897	0.706	0.858
	BERT model [4]	0.977	0.967	0.975
	Prompt model	0.980	0.975	0.978
i2b2 2012	Logistic Regression	0.921	0.782	0.893
	NegEx [5]	0.937	0.815	0.917
	RadText [35]	0.898	0.607	0.853
	BERT model [4]	0.955	0.866	0.940
	Prompt model	0.956	0.875	0.943
BioScope	Logistic Regression	0.945	0.780	0.914
	NegEx [5]	0.879	0.621	0.847
	RadText [35]	0.836	0.631	0.789
	BERT model [4]	0.951	0.835	0.928
	Prompt model	0.966	0.823	0.938
MIMIC-III	Logistic Regression	0.899	0.846	0.883
	NegEx [5]	0.908	0.863	0.896
	RadText [35]	0.880	0.700	0.890
	BERT model [4]	0.951	0.937	0.947
	Prompt model	0.950	0.933	0.950
NegEx	Logistic Regression	0.926	0.821	0.889
	NegEx [5]	0.983	0.931	0.972
	RadText [35]	0.817	0.530	0.734
	BERT model [4]	0.926	0.815	0.890
	Prompt model	0.940	0.821	0.938
Chia	Logistic Regression	0.693	0.540	0.609
	NegEx [5]	0.763	0.612	0.705
	RadText [35]	0.703	0.430	0.609
	BERT model [4]	0.763	0.619	0.708
	Prompt model	0.772	0.652	0.724