Supporting the annotation of chronic obstructive pulmonary disease (COPD) phenotypes with text mining workflows

Table 6 Results of evaluation using a fixed split over 381 paragraphs (training set: 75% or 286 paragraphs; held-out set: 25% or 95 paragraphs), using exact matching

	Concept recognisers currently in Argo			Concept recognisers trained on our corpus
	Precision	Recall	F-score	Precision	Recall	F-score
AnatomicalConcept	0.2602	0.6145	0.3656	0.8000	0.4314	0.5605
Drug	0.6885	0.1900	0.2979	0.7966	0.4196	0.5497
MedicalCondition	0.4494	0.2492	0.3206	0.8673	0.3899	0.5380
TestOrMeasure	0.0250	0.0041	0.0070	0.6719	0.2966	0.4115
Treatment	0.4111	0.0847	0.1404	0.8400	0.2903	0.4315
Micro-average	0.3735	0.1614	0.2254	0.8034	0.3552	0.4926
Macro-average	0.3669	0.2285	0.2816	0.7952	0.3656	0.5009

ISSN: 2041-1480