期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2017
卷号：95
期号：12
页码：2814
出版社：Journal of Theoretical and Applied
摘要：One of the areas in text mining which is text classification has attracted much attention in various industries and fields lately. This is because the text classification has the ability in labelling text documents to one or more pre-defined categories based on content similarity. As text classification emphasizes on document level, question classification works at finer level such as sentence and phrase. Several studies on question classification in respect to Bloom taxonomy to measure cognitive level of learners in higher learning institutions have been carried out in the past. But, existing feature types in the past work may work reasonably well on data sets consisting of questions that are too specific to one particular field or area which will result in having multiple classifiers to be built for questions involving various fields or areas. Certainly, feature types play an important role in improving the accuracy of classifier. Past related work emphasizes on feature types such as bag of word (BOW) and syntactic analysis in question classification. In this study, a new feature type named taxonomy based is proposed to improve the accuracy of question classification for data sets having questions from various fields. The performance of question classification using the new feature type between data sets consisting of questions from specific and various areas will be compared. Support Vector Machine classifier will be used as it is known for high accuracy in text classification. The outcome of this study shows that the taxonomy based features has the ability in improve the accuracy of classifier involving data sets of questions from various fields
关键词：Feature Type; Question Classification; Support Vector Machine; Bloom Taxonomy; Bag-Of-Words
