Table 5.

NER performances on synthetic and natural clinical corpora: natural_for_synthetic, natural_for_external_1, and natural_for_external_2 refer to models trained from natural and tested on synthetic, external_1, and external_2 test datasets, respectively

Entity	natural_for_synth			natural_for_external_1			natural_for_external_2
	Precision	Recall	F1	Precision	Recall	F1	Precision	Recall	F1
Problem	0.772 (0.866)	0.841 (0.950)	0.805 (0.906)	0.703 (0.849)	0.749 (0.920)	0.725 (0.883)	0.707 (0.820)	0.800 (0.939)	0.751 (0.875)
Test	0.768 (0.841)	0.762 (0.841)	0.764 (0.840)	0.605 (0.703)	0.682 (0.803)	0.640 (0.749)	0.726 (0.818)	0.736 (0.835)	0.730 (0.826)
Treatment	0.749 (0.852)	0.787 (0.897)	0.767 (0.874)	0.658 (0.791)	0.666 (0.827)	0.662 (0.808)	0.660 (0.778)	0.750 (0.914)	0.701 (0.838)
Overall	0.765 (0.858)	0.812 (0.915)	0.788 (0.886)	0.686 (0.825)	0.728 (0.892)	0.706 (0.857)	0.700 (0.810)	0.778 (0.915)	0.737 (0.859)

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