Figure 10.

Comparison of precision, recall, and F1-measure results achieved on AnEM and NCBI corpora for named entity recognition, considering exact (E), left (L), right (R), shared (S) and overlap (O) matching. The various sub-classes from each corpus were merged into a single class, in order to evaluate the general ability to recognize disorder and anatomical concept names.

Campos et al. BMC Bioinformatics 2013 14:281   doi:10.1186/1471-2105-14-281
Download authors' original image