Sentiment Analysis of Student Opinions in Large-scale Open Online Courses Using Automatic Machine Learning Techniques: What does it tell us?

Khe Foon Hew, Chen Qiao, Yumeng Sun and Ying Tang
The Unviersity of Hong Kong
Hong Kong SAR, China

Student opinions play a very important role in education — they can influence student behaviours, such as whether or not to pay attention, which in turn influences their decision to drop out or continue learning. This study offers a new contribution by using sentiment analysis, otherwise known as “opinion mining”, a technique for analyzing and classifying sentiments found in a large-scale corpus of reflective sentences (75,239 sentences) posted by 18,032 students who completed one or more of 218 MOOCs. The open-sourced text processing package TextBlob was employed as the sentiment analysis engine, which computes text sentiment by averaging the term “sentiments” of the text based on a sentiment dictionary derived from WordNet3. We explored and described the students’ positive and negative sentiments with respect to one or more of the following six aspects: (a) structure and pace; (b) video; (c) instructor; (d) content and resources; (e) interaction and support; and (f) assignment and assessment.