Skip to Main Content
Table 4: 

Results on the original CCGbank evaluation set (WSJ section 23). The population n for computing Parseability is the number of sentences in the test set. In each column, we highlight all results that fall within the standard deviation of the best result.

ModelAccAcc by cat freq in trainingParsing
All≥10010–991–9OOVLFParseability
n =55,371n =54,825n =442n =82n =22n =2,407
N=435N=171N=176N=67N=21
Nonconstructive 
MLP_10@ 96.09 ± .07 96.50 ± .08 67.27 ± 1.02 – – 90.78 ± .09 86.95 ± 0.75 
MLP_1 96.22 ± .06 96.58 ± .07 70.29 ± 2.35 23.17 ± 3.23 – 90.91 ± .09 88.26 ± 0.39 
 
Constructive 
K+19 92.12 ± .21 92.46 ± .20 65.38 ± 0.99 34.55 ± 4.28 1.52 ± 2.62 87.66 ± .19 91.14 ± 0.13 
RNN@ 95.10 ± .07 95.48 ± .07 65.76 ± 1.71 26.02 ± 0.70 0.00 ± 0.00 90.63 ± .04 89.53 ± 0.18 
AddrMLP@ 96.09 ± .07 96.44 ± .08 68.10 ± 1.38 37.40 ± 1.41 3.03 ± 2.62 90.79 ± .08 86.03 ± 1.72 
ModelAccAcc by cat freq in trainingParsing
All≥10010–991–9OOVLFParseability
n =55,371n =54,825n =442n =82n =22n =2,407
N=435N=171N=176N=67N=21
Nonconstructive 
MLP_10@ 96.09 ± .07 96.50 ± .08 67.27 ± 1.02 – – 90.78 ± .09 86.95 ± 0.75 
MLP_1 96.22 ± .06 96.58 ± .07 70.29 ± 2.35 23.17 ± 3.23 – 90.91 ± .09 88.26 ± 0.39 
 
Constructive 
K+19 92.12 ± .21 92.46 ± .20 65.38 ± 0.99 34.55 ± 4.28 1.52 ± 2.62 87.66 ± .19 91.14 ± 0.13 
RNN@ 95.10 ± .07 95.48 ± .07 65.76 ± 1.71 26.02 ± 0.70 0.00 ± 0.00 90.63 ± .04 89.53 ± 0.18 
AddrMLP@ 96.09 ± .07 96.44 ± .08 68.10 ± 1.38 37.40 ± 1.41 3.03 ± 2.62 90.79 ± .08 86.03 ± 1.72 
Close Modal

or Create an Account

Close Modal
Close Modal