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

Table 7 Precision/Recall/F1-score results for gene mention over CRAFT development set: BANNER with distributed model trained on BioCreative II using indicated evaluation mapping strategies

	BANNER BioCreative protein-STAR			BANNER BioCreative protein-GENESTAR			BANNER BioCreative protein-POLYSTAR
	Prec	Recall	F1	Prec	Recall	F1	Prec	Recall	F1
strict	0.38	0.61	0.47	0.16	0.49	0.25	0.20	0.78	0.32
overlap	0.49	0.80	0.61	0.25	0.77	0.38	0.22	0.85	0.35
shared	0.49	0.79	0.60	0.25	0.75	0.37	0.22	0.85	0.35
subspan	0.49	0.80	0.61	0.25	0.76	0.38	0.22	0.85	0.35

ISSN: 1471-2105