... statis-tical feature selection. The proposed method dealsonly with those features that are statistically signif-icant for kernel calculation, large significant sub-structures can be used without ... effi-ciently, regardless of its size, to solve NLP tasks.4 Proposed Feature Selection MethodOur approach is basedon statisticalfeature selection in contrast to the conventional methods, which usesub-structure ... features, is quite natural. We note, however,that kernels are calculated using the DP algorithm.Therefore, it is not clear how to calculate kernels ef-ficiently with a statistical feature selection...