Kernels for the Classification of Questions in Spanish & Catalan

Work on the automatic classification of questions for Spanish & Catalan question answering systems is reported. After explaining what a kernel & kernel method are, a system based on the SVM (Support Vector Machines) algorithm is constructed to identify various types of definitional kernels in natural language processing. Corpus construction on the basis of the English-language TREC, question taxonomy, & vector coding are discussed. Four experiments were conducted to test: (1) the influence of kernel parameters, (2) precision degree between kernels, (3) comparison of characteristics vectors, & (4) system performance in each of the two languages. The essay concludes with a brief survey of literature reporting research & work on question answering systems, mostly in English, & outline of plans for future work on improving the system described here. Adapted from the source document