Detecting concept mentions in biomedical text using hidden Markov model: multiple concept types at once or one at a time?

Table 2 Comparison of detection performance

			TP	FP	FN	Prec.	Rec.	F-score
i2b2/VA	Problem	All-at-once	964	267	231	0.783	0.806	0.794
	Problem	One-at-a-time	932	244	264	0.792	0.779	0.785
	Test	All-at-once	582	114	153	0.835	0.791	0.813
	Test	One-at-a-time	551	112	185	0.831	0.748	0.787
	Treatment	All-at-once	653	139	196	0.823	0.769	0.795
	Treatment	One-at-a-time	625	138	223	0.818	0.737	0.775
JNLPBA	Protein	All-at-once	2,373	840	653	0.739	0.784	0.761
	Protein	One-at-a-time	2,251	752	775	0.749	0.744	0.747
	DNA	All-at-once	581	270	371	0.683	0.610	0.644
	DNA	One-at-a-time	527	339	425	0.609	0.553	0.580
	Cell Type	All-at-once	496	167	174	0.748	0.740	0.744
	Cell Type	One-at-a-time	455	168	215	0.730	0.678	0.703
	Cell Line	All-at-once	233	102	149	0.695	0.610	0.649
	Cell Line	One-at-a-time	212	180	170	0.543	0.554	0.548
	RNA	All-at-once	36	24	59	0.594	0.383	0.462
	RNA	One-at-a-time	33	18	62	0.640	0.345	0.447

ISSN: 2041-1480