Let us mention the popularity of encoding texts into numerical vectors, and the proposal and the application of string kernels as the solution to the above problems. In 2002, Sebastiani presented the numerical vectors are the standard representations of texts in applying the machine learning algorithms to the text classifications [4]. In 2002, Lodhi et al. proposed the string kernel as a kernel function of raw texts in using the SVM (Support Vector Machine) to the text classification [5]. In 2004, Lesile et al. used the version of SVM which proposed by Lodhi et al. to the protein classification [6]. In 2004, Kate and Mooney used also the SVM version for classifying sentences by their meanings [7].