Table 2.

Performance of relation extraction on the test set (best precision, recall, and F1 score are highlighted in bold)

Model	Performance (lenient/relaxed)a
	precision	recall	F1 score
SVMs-1b	0.9623	0.9300	0.9459
SVMs-4	0.9605	0.9422	0.9512
RFs-4	0.9612	0.9350	0.9479
GB-4	0.9730	0.9541	0.9635

Abbreviations: GB, gradient boosting; RFs: random forests; SVMs, support vector machines; .

^aThe lenient score and relaxed score are the same for subtask 2.

^bSVMs-1 is the final relation extraction system submitted during this challenge (ranked fourth).