Table 4.

NER performances on synthetic and natural clinical corpora: synthetic_for_natural, synthetic_for_external_1, and synthetic_for_external_2 refer to models trained from synthetic and tested on natural, external_1, and external_2 test datasets, respectively

Entity	synth_for_natural			synth_for_external_1			synth_for_external_2
	Precision	Recall	F1	Precision	Recall	F1	Precision	Recall	F1
Problem	0.807 (0.930)	0.784 (0.899)	0.795 (0.914)	0.705 (0.842)	0.752 (0.909)	0.727 (0.874)	0.750 (0.869)	0.773 (0.904)	0.761 (0.886)
Test	0.825 (0.911)	0.724 (0.793)	0.770 (0.848)	0.674 (0.807)	0.642 (0.768)	0.657 (0.786)	0.780 (0.882)	0.750 (0.852)	0.763 (0.865)
Treatment	0.802 (0.893)	0.782 (0.872)	0.791 (0.882)	0.682 (0.830)	0.632 (0.770)	0.656 (0.798)	0.685 (0.811)	0.708 (0.858)	0.695 (0.832)
Overall	0.809 (0.918)	0.772 (0.872)	0.790 (0.894)	0.698 (0.837)	0.720 (0.871)	0.709 (0.854)	0.741 (0.859)	0.755 (0.884)	0.748 (0.871)

Numbers in the parentheses are results based on relaxed matching criteria.