Table 3.

Text mining evaluation, showing performance results on the NLM-Gene test set, when the training is performed on the listed combinations of training sets. The results show the effect on performance improvement due to the NLM-Gene corpus, and specific system upgrades.

Model		Gene name entity recognition			Gene entity normalization
	Train	Precision	Recall	f-measure	Precision	Recall	f-measure
GNormPlus (original)	GN	0.919	0.762	0.833	0.714	0.619	0.663
GNormPlus (original)	GN + NLM	0.922	0.825	0.871	0.724	0.697	0.710
GNormPlus (original) + Deep Learning	GN + NLM	0.929	0.821	0.872	0.726	0.697	0.711
GNormPlus (upgraded)	GN + NLM	0.933	0.834	0.881	0.748	0.707	0.727
GNormPlus (upgraded) + (species insensitive)	GN + NLM	0.933	0.834	0.881	0.879	0.840	0.859