Using contextual and lexical features to restructure and validate the classification of biomedical concepts

BMC Bioinformatics

Table 1 Error rates of the experiments

	The entire test set (N = 223)		≥ 1 syntactic dependencies (N = 192)		≥ 10 syntactic dependencies (N = 91)
Type of features	Error rate	MRR	Error rate	MRR	Error rate	MRR
Syntactic dependencies	--	--	0.315	0.792	0.187	0.887
Strings without annotations*	0.191	0.881	0.221^#	0.862^#	0.247^#	0.853^#
Strings with annotations*	0.143	0.907	0.164^#	0.895^#	0.192^#	0.886^#

* Note that the string-based approach is independent of the number of available syntactic features used by the distributional approach, and the alignment (^#) is just for comparison purpose.

ISSN: 1471-2105