Table 4.

Evaluation against the CDR corpus

	Precision	Recall	F 1
SemRep-ALL	0.90	0.24	0.38
SemRep-SENTENCE	0.90	0.35	0.50
Xu et al. [19]	0.56	0.58	0.57
Peng et al. [20]	0.66	0.57	0.61

SemRep-ALL indicates the case in which all ground truth relations are taken into account. SemRep-SENTENCE indicates the scenario in which only the intra-sentence ground truth relations are considered. Xu et al. [19] was the top-ranking system in the BioCreative V CID task and Peng et al. [20] reported best post-challenge results. Both systems perform end-to-end relation extraction