The frequency and P / R / F1 scores for typical argument role labels on the CoNLL-2009 English test set. Baseline refers to the tree-based model, and +Pruning refers to the SynRule Soft Pruning method. Due to the width constraint, we omit the Precision (P) and Recall (R) scores of the nominal pred.
| . | verbal pred . | nominal pred . | ||||
|---|---|---|---|---|---|---|
| Role . | FREQ . | Baseline . | +Pruning . | FREQ . | Baseline . | +Pruning . |
| A0 | 15% | 93.1 / 91.9 / 92.5 | 93.2 / 92.3 / 92.7 | 10% | 83.3 | 83.6 |
| A1 | 21% | 93.9 / 93.1 / 93.5 | 93.6 / 93.4 / 93.5 | 16% | 87.2 | 87.4 |
| A2 | 5% | 84.3 / 81.8 / 83.0 | 85.1 / 83.1 / 84.1 | 7% | 81.0 | 82.8 |
| AM-* | 16% | 82.2 / 80.2 / 81.2 | 82.4 / 80.6 / 81.5 | 5% | 75.4 | 76.3 |
| . | verbal pred . | nominal pred . | ||||
|---|---|---|---|---|---|---|
| Role . | FREQ . | Baseline . | +Pruning . | FREQ . | Baseline . | +Pruning . |
| A0 | 15% | 93.1 / 91.9 / 92.5 | 93.2 / 92.3 / 92.7 | 10% | 83.3 | 83.6 |
| A1 | 21% | 93.9 / 93.1 / 93.5 | 93.6 / 93.4 / 93.5 | 16% | 87.2 | 87.4 |
| A2 | 5% | 84.3 / 81.8 / 83.0 | 85.1 / 83.1 / 84.1 | 7% | 81.0 | 82.8 |
| AM-* | 16% | 82.2 / 80.2 / 81.2 | 82.4 / 80.6 / 81.5 | 5% | 75.4 | 76.3 |