Skip to Main Content
Table 4 

Accuracy rates on TOEFL11 corpus (English L2) of various classification systems based on string kernels compared with other state-of-the-art approaches. The best accuracy rates on each set of experiments are highlighted in bold. The weights a1 and a2 from the weighted sums of kernels are computed by kernel alignment.

MethodDev10-fold CVTest
Ensemble model (Tetreault et al. 2012) 80.9% 
KRR and string kernels (Popescu and Ionescu 2013) 82.6% 82.7% 
SVM and word features (Jarvis, Bestgen, and Pepper 2013) 84.583.6% 
Ensemble model and CFG (Bykh and Meurers 2014) 84.8% 
KRR and  85.4% 82.5% 82.0% 
KRR and  84.9% 82.2% 82.6% 
KRR and  78.7% 77.1% 77.5% 
KRR and  85.7% 82.6% 82.7% 
KRR and  84.9% 82.2% 82.0% 
KRR and  85.5% 82.6% 82.5% 
KRR and  85.5% 82.6% 82.5% 
 
KDA and  86.2% 83.6% 83.6% 
KDA and  85.2% 83.5% 84.6% 
KDA and  79.7% 78.5% 79.2% 
KDA and  87.184.0% 84.7% 
KDA and  85.8% 83.4% 83.9% 
KDA and  86.4% 84.1% 85.0% 
KDA and  86.5% 84.1% 85.3
KDA and  87.0% 84.1% 84.8% 
MethodDev10-fold CVTest
Ensemble model (Tetreault et al. 2012) 80.9% 
KRR and string kernels (Popescu and Ionescu 2013) 82.6% 82.7% 
SVM and word features (Jarvis, Bestgen, and Pepper 2013) 84.583.6% 
Ensemble model and CFG (Bykh and Meurers 2014) 84.8% 
KRR and  85.4% 82.5% 82.0% 
KRR and  84.9% 82.2% 82.6% 
KRR and  78.7% 77.1% 77.5% 
KRR and  85.7% 82.6% 82.7% 
KRR and  84.9% 82.2% 82.0% 
KRR and  85.5% 82.6% 82.5% 
KRR and  85.5% 82.6% 82.5% 
 
KDA and  86.2% 83.6% 83.6% 
KDA and  85.2% 83.5% 84.6% 
KDA and  79.7% 78.5% 79.2% 
KDA and  87.184.0% 84.7% 
KDA and  85.8% 83.4% 83.9% 
KDA and  86.4% 84.1% 85.0% 
KDA and  86.5% 84.1% 85.3
KDA and  87.0% 84.1% 84.8% 
Close Modal

or Create an Account

Close Modal
Close Modal