Table 8.

Class based performance (F-score in %) for BioCreative corpus using various tokenizers.

Algorithm	Entity type	Development set			Test set
		ChemSpot	tmVar	ChemTok	ChemSpot	tmVar	ChemTok
CRF	Abbreviation	68.14	66.58	68.68	67.20	65.42	68.86
	Family	69.22	67.59	72.67	71.94	70.60	74.91
	Formula	76.57	69.70	80.12	75.29	69.81	79.37
	Identifier	63.03	59.45	74.60	63.88	61.55	74.28
	Multiple	32.50	26.86	41.35	32.77	30.50	35.12
	Systematic	79.41	78.09	82.85	79.95	78.33	83.11
	Trivial	85.62	84.11	88.36	85.52	83.69	87.66
SVM	Abbreviation	72.59	72.12	74.50	72.42	71.65	73.75
	Family	69.82	69.69	71.99	71.81	71.57	74.31
	Formula	82.667	81.61	84.68	82.15	81.68	85.28
	Identifier	72.08	69.76	72.52	74.60	74.76	77.18
	Multiple	36.06	34.31	39.26	26.89	20.96	30.46
	Systematic	82.33	81.51	84.73	82.10	81.49	84.44
	Trivial	86.73	85.85	88.84	86.50	86.14	88.51