Table 2 F scores (at confidence threshold of 0.3) and Mean Average Precision (MAP) values for Figs. 1–5.
From: Cascaded classifiers for confidence-based chemical named entity recognition
Corpus | System | MAP | F |
|---|---|---|---|
Chemistry | Full | 87.1% | 80.8% |
Chemistry | No Rescorer | 86.8% | 81.0% |
Chemistry | No Preclassifier | 82.7% | 74.8% |
Chemistry | No n-Grams | 79.2% | 72.2% |
Chemistry | Custom LingPipe | 75.9% | 74.6% |
Chemistry | Pure LingPipe | 66.9% | 63.2% |
Chemistry | No Overlaps | 82.9% | 80.8% |
Chemistry | CM | 87.0% | 81.2% |
Chemistry | RN | 74.5% | 73.4% |
Chemistry | CJ | 90.0% | 92.0% |
Chemistry | ASE | 17.4% | 36.2% |
PubMed | Full | 86.1% | 83.2% |
PubMed | No Rescorer | 83.3% | 79.1% |
PubMed | No Preclassifier | 81.4% | 73.4% |
PubMed | No n-Grams | 77.6% | 70.6% |
PubMed | Custom LingPipe | 78.6% | 75.6% |
PubMed | Pure LingPipe | 71.9% | 66.1% |
PubMed | CM | 85.6% | 82.3% |
PubMed | RN | 95.3% | 93.2% |
PubMed | CJ | 78.7% | 83.1% |
PubMed | ASE | 83.4% | 86.0% |