A corpus of full-text journal articles is a robust evaluation tool for revealing differences in performance of biomedical natural language processing tools

Table 6 Precision/Recall/F1-score results for gene mention over CRAFT initial release set: ABNER with distributed model trained on NLPBA using indicated evaluation mapping strategies

	ABNER NLPBA protein-STAR			ABNER NLPBA star-POLYSTAR			ABNER NLPBA protein-GENESTAR
	Prec	Recall	F1	Prec	Recall	F1	Prec	Recall	F1
strict	0.30	0.34	0.32	0.11	0.41	0.17	0.13	0.29	0.18
overlap	0.39	0.44	0.41	0.14	0.57	0.23	0.15	0.36	0.22
shared	0.38	0.42	0.40	0.14	0.54	0.22	0.15	0.36	0.21
subspan	0.38	0.43	0.41	0.14	0.57	0.23	0.15	0.36	0.22

ISSN: 1471-2105