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

Table 12 Precision/Recall/F1-score results for gene mention over CRAFT initial release set: LingPipe with distributed model trained on GeneTag using indicated evaluation mapping strategies

	LingPipe GeneTag gene-STAR			LingPipe GeneTag gene-GENE			LingPipe GeneTag gene-GENESTAR
	Prec	Recall	F1	Prec	Recall	F1	Prec	Recall	F1
strict	0.22	0.63	0.33	0.08	0.56	0.15	0.10	0.58	0.17
shared	0.30	0.85	0.44	0.12	0.84	0.22	0.14	0.84	0.24
subspan	0.30	0.86	0.45	0.13	0.86	0.22	0.14	0.86	0.24
overlap	0.30	0.86	0.45	0.13	0.87	0.22	0.14	0.87	0.25

ISSN: 1471-2105