Automated Evaluation of Student Comments on Their Learning Behavior

摘要

Learning comments are valuable sources of interpreting student status of understanding. The PCN method introduced in [Gouda2011] analyzes the attitudes of a student from a view point of time series. Each sentence of a comment is manually classified as one of P,C,N or O sentence. P(previous) indicates learning activities before the classtime, C(current) represents understanding or achievements during the classtime, and N(next) means a learning activity plan or goal until next class. The present paper applies SVM(Support Vecotor Machine) to predict the category to which a given sentence belongs. Empirical evaluation using 4,086 sentences was conducted. By selecting feature words of each category, the prediction performance was satisfactory with F-measures 0.8203, 0.7352, 0.8416 and 0.8612 for P,C,N and O respectively.