Table 23 Results of constituent parsers using retrained CRAFT models for each CRAFT fold and the development set compared to untrained results on the development set; labeled bracket precision (LB-P), recall (LB-R) and F-score (LB-F)
Parser | Fold 0 | Fold 1 | Fold 2 | Fold 3 | Fold 4 | Training Average | Dev Set | Dev Set Untrained |
|---|---|---|---|---|---|---|---|---|
Berkeley | ||||||||
LB-P | 82.75 | 92.02 | 84.63 | 83.70 | 83.85 | 85.39 | 83.98 | 61.60 |
LB-R | 82.64 | 90.82 | 84.01 | 83.29 | 82.88 | 84.73 | 83.20 | 64.50 |
LB-F | 82.70 | 91.41 | 84.32 | 83.49 | 83.36 | 85.06 | 83.59 | 63.02 |
Bikel | ||||||||
LB-P | 80.49 | 81.10 | 81.18 | 80.77 | 91.43 | 82.99 | 80.86 | 63.97 |
LB-R | 79.68 | 79.77 | 80.10 | 80.46 | 91.06 | 82.21 | 80.44 | 65.82 |
LB-F | 80.08 | 80.43 | 80.64 | 80.62 | 91.24 | 82.60 | 80.65 | 64.89 |
Stanford 1.6.6 | ||||||||
LB-P | 75.65 | 75.86 | 77.71 | 76.21 | 77.86 | 76.65 | 76.17 | 60.76 |
LB-R | 76.81 | 76.84 | 78.65 | 77.24 | 77.85 | 77.48 | 75.92 | 64.70 |
LB-F | 76.23 | 76.34 | 78.18 | 76.72 | 77.86 | 77.07 | 76.04 | 62.67 |