Table 5 Evaluation results of the approach on cross-corpus non-informative sentence-removed datasets (we report percentages)
From: Biomedical named entity extraction: some issues of corpus compatibilities
Approach | Training set | Test set | r | p | FM |
|---|---|---|---|---|---|
Best Individual Classifier | JNLPBA (protein only)+AIMed | AIMed | 80.58 | 84.43 | 82.46 |
SOO Based Ensemble | JNLPBA (protein only)+AIMed | AIMed | 81.98 | 86.01 | 83.95 |
Best Individual Classifier | JNLPBA (protein + DNA)+AIMed | AIMed | 84.66 | 83.50 | 84.08 |
SOO Based Ensemble | JNLPBA (protein + DNA)+AIMed | AIMed | 86.07 | 85.01 | 85.54 |
Best Individual Classifier | JNLPBA (protein only)+GENETAG | GENETAG | 91.79 | 90.61 | 91.20 |
SOO Based Ensemble | JNLPBA (protein only)+GENETAG | GENETAG | 93.19 | 92.08 | 92.63 |
Best Individual Classifier | JNLPBA (protein + DNA + RNA)+GENTAG | GENTAG | 93.98 | 90.67 | 92.29 |
SOO Based Ensemble | JNLPBA (protein + DNA + RNA)+GENTAG | GENTAG | 95.09 | 92.16 | 93.60 |