Table 6.

NER performances on natural+synthetic and natural corpora: natural+synthetic_for_natural and natural_for_natural refer to models trained from the augmented corpus of natural+synthetic and the natural corpus (training set), respectively, and then tested on natural test set

Entity	natural+synthetic_for_natural			natural_for_natural
	Precision	Recall	F1	Precision	Recall	F1
Problem	0.862 (0.948)	0.848 (0.930)	0.855 (0.939)	0.825 (0.921)	0.841 (0.938)	0.833 (0.929)
Test	0.857 (0.913)	0.850 (0.899)	0.853 (0.905)	0.838 (0.901)	0.821 (0.877)	0.829 (0.888)
Treatment	0.849 (0.930)	0.828 (0.909)	0.838 (0.919)	0.828 (0.914)	0.794 (0.881)	0.810 (0.896)
Overall	0.858 (0.936)	0.844 (0.919)	0.851 (0.927)	0.829 (0.914)	0.827 (0.913)	0.828 (0.914)

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