Table 2. Summary benchmark of LINNAEUS and SPECIES.
| Corpus | Level | Software | Precision | Recall | F1 |
| S800 | Document | LINNAEUS | 86.4% | 89.3% | 87.9% |
| SPECIES | 85.9% | 89.8% | 87.8% | ||
| Mention | LINNAEUS | 84.3% | 75.4% | 79.6% | |
| SPECIES | 83.9% | 72.6% | 77.8% | ||
| L100E | Document | LINNAEUS | 89.2% | 91.4% | 90.3% |
| SPECIES | 89.9% | 94.3% | 92.0% | ||
| Mention | LINNAEUS | 88.7% | 81.8% | 85.1% | |
| SPECIES | 91.5% | 90.8% | 91.1% |
We compared LINNAEUS and SPECIES taggers by calculating their precision and recall on two different corpora (L100E an S800) at the document and at the mention level.
Unsurprisingly, SPECIES performs better than LINNAEUS on the L100E corpus, which we used during the development SPECIES. On the S800 corpus, which did not exist when either tagger was developed, we obtain very similar performance numbers for the two taggers.