Table 2 |
||
| The tagging performance of BANNER and OSCAR3 | ||
| Protein names tagged | Small molecule names | |
| by BANNER | tagged by OSCAR3 | |
| Pantothenate and coenzyme A biosynthesis pathway | ||
| Recall(C) (%) | 81 (112/139) | 96 (329/343) |
| Precision (%) | 85 (112/132) | 86 (329/384) |
| F-score (%) | 83 | 91 |
| Tetrahydrofolate biosynthesis pathway | ||
| Recall(C) (%) | 93 (250/268) | 82 (528/647) |
| Precision (%) | 76 (250/327) | 95 (528/558) |
| F-score (%) | 84 | 88 |
| Aerobic fatty acid β-oxidation I pathway | ||
| Recall(C) (%) | 91 (341/376) | 81 (456/565) |
| Precision (%) | 82 (341/414) | 92 (456/494) |
| F-score (%) | 86 | 86 |
The tagging performance of the NER tools when applied to the Abstracts and Introductions from papers referenced in EcoCyc with respect to our three evaluation pathways. Taking the BANNER column for the pantothenate and coenzyme A biosynthesis pathway as an example, the numbers in brackets indicate that BANNER correctly identified 112 out of the 139 protein names (recall row); and of the 132 names it tagged, 112 were correct (precision row). The OSCAR3 results are with a confidence threshold of zero.
Czarnecki et al. BMC Bioinformatics 2012 13:172 doi:10.1186/1471-2105-13-172