Table 2.

Evaluation results for the NCB14 dataset.

	Textractor				NLProt				Union
PubMedID	Correct	Incorrect	Precision	Rel. Recall	Correct	Incorrect	Precision	Rel. Recall	Correct
12629548	41	19	68%	67%	37	77	32%	59%	63
12629549	38	3	93%	84%	32	79	29%	40%	50
12640462	23	17	58%	100%	1	13	7%	4%	23
12640463	18	17	51%	36%	46	77	37%	87%	53
12640464	67	18	79%	72%	50	26	66%	50%	100
12669071	14	8	64%	54%	19	35	35%	73%	26
12669072	14	3	82%	65%	18	27	40%	51%	26
12669073	27	1	96%	81%	25	31	45%	53%	39
12669075	19	7	73%	73%	16	26	38%	47%	26
12669077	19	5	79%	73%	16	38	30%	39%	28
12669082	86	26	77%	68%	82	87	49%	65%	126
12669083	64	13	83%	58%	81	74	52%	72%	112
12679784	25	7	78%	58%	26	37	41%	60%	43
12692559	54	26	68%	60%	67	77	47%	55%	99
micro-evaluation	509	170	75%	63%	516	704	42%	63%	814
macro-evaluation			75%	67%			39%	54%