Table 1.

LocText (RE only) and STRING Tagger (NER); intrinsic evaluation

Method and evaluation	P	R	F ±StdErr
STRING Tagger Total	84%	78%	81% ± 1
STRING Tagger on Protein	80%	78%	79% ± 2
STRING Tagger on Location	90%	71%	80% ± 3
STRING Tagger on Organism	96%	92%	94% ± 1
LocText, with manual entities	93%	68%	79% ± 3
Baseline, with manual entities	75%	74%	74% ± 3

Performances of the NER and RE components independently evaluated on the LocTextCorpus; P=precision, R=recall, F ±StdErr=F-measure with standard error