Chinese test set results (PCTB articles 271–300) applying sentence-level high-precision and unary constraints to two parsers with parameter settings tuned on development data.
Parser . | F1 . | Precision . | Recall . | Seconds . | Speed-up . |
---|---|---|---|---|---|
BUBS (2010) | 79.5 | 79.5 | 79.1 | 169 | |
+ Unary(50) | 80.7 | 82.1 | 79.4 | 153 | 1.1x |
+ HP(0.8) | 81.1 | 81.5 | 80.7 | 75 | 2.3x |
+ HP(0.8) + Unary(50) | 81.8 | 83.1 | 80.5 | 44 | 3.8x |
Berkeley (2007) | 83.9 | 84.5 | 83.3 | 141 | |
+ Unary(50) | 84.5 | 85.9 | 83.0 | 125 | 1.1x |
+ HP(0.7) | 84.5 | 85.1 | 83.8 | 64 | 2.2x |
+ HP(0.7) + Unary(50) | 84.7 | 86.1 | 83.4 | 57 | 2.5x |
Parser . | F1 . | Precision . | Recall . | Seconds . | Speed-up . |
---|---|---|---|---|---|
BUBS (2010) | 79.5 | 79.5 | 79.1 | 169 | |
+ Unary(50) | 80.7 | 82.1 | 79.4 | 153 | 1.1x |
+ HP(0.8) | 81.1 | 81.5 | 80.7 | 75 | 2.3x |
+ HP(0.8) + Unary(50) | 81.8 | 83.1 | 80.5 | 44 | 3.8x |
Berkeley (2007) | 83.9 | 84.5 | 83.3 | 141 | |
+ Unary(50) | 84.5 | 85.9 | 83.0 | 125 | 1.1x |
+ HP(0.7) | 84.5 | 85.1 | 83.8 | 64 | 2.2x |
+ HP(0.7) + Unary(50) | 84.7 | 86.1 | 83.4 | 57 | 2.5x |