Learning Sentiment from Students' Feedback for Real-Time Interventions in Classrooms

Knowledge about users sentiments can be used for a variety of adaptation purposes. In the case of teaching, knowledge about students sentiments can be used to address problems like confusion and boredom which affect students engagement. For this purpose, we looked at several methods that could be used for learning sentiment from students feedback. Thus, Naive Bayes, Complement Naive Bayes (CNB), Maximum Entropy and Support Vector Machine (SVM) were trained using real students' feedback. Two classifiers stand out as better at learning sentiment, with SVM resulting in the highest accuracy at 94%, followed by CNB at 84%. We also experimented with the use of the neutral class and the results indicated that, generally, classifiers perform better when the neutral class is excluded.