Automatic Classification of Questions into Bloom's Cognitive Levels Using Support Vector Machines
Anwar Ali Yahya, Addin Osama · 2011
In recent years, E-learning has increasingly become a promising technology in educational institutions. Among numerous components of E-learning systems, question bank is a primordial component. Question bank is a repository of questions that assists students and instructors in the educational process. In question bank, questions are annotated, stored and retrieved based on predefined criteria such as bloom's cognitive levels. Definitely, for question bank management, the automatic classification of questions according to Bloom's cognitive levels is of particular benefit. This paper explores the effectiveness of support vector machines (SVMs), in tackling the problem of question classification into Bloom's cognitive levels. To do so, a dataset of pre-classified questions has been collected. Each question is processed through removal of punctuations and stop words, tokenization, stemming, term weighting and length normalization. SVMs classifiers, namely linear kernel, have been built and evaluated on approximately 70% and 30% of the dataset respectively, using SVM-Light software package. The obtained preliminary results show a satisfactory effectiveness of SVMs with respect to classification accuracy and precision. However, due to the small size of the current dataset, the results of the classifiers' recall and F-measure suggest a need for further experiments with larger dataset to obtain conclusive results.