Figure 5

The three diagrams show the evaluation of the different tagging solutions against the BioCreative-II corpus only. For the evaluation represented by the solid line only the gene list of the BioCreative-II evaluation kit has been used. For the two other lines also the alternative gene list has been applied counting the '*TP’ (dotted line) and the '**TP’ (dashed line) as TP candidates (instead of FP candidates). From the first to the last evaluation all three measures improve, and only the ML-Tag solutions profit from the '**TP’ data entries in the alternative gene list of BioCreative-II. Obviously these candidates are not part of any lexical resource and only occur in the BioCreative-II corpus.