Educational Decision Support System Adopting Sentiment Analysis on Student Feedback

Educational institutions are constantly analyzing their teaching practice and learning environments to provide a better learning experience for their students. Engaging with all students’ feedback and analyzing manually is almost impossible due to the amount of textual data. Sentiment analysis has the potential to analyze students’ feedback and extract their opinion or sentiment toward courses, teaching, and infrastructure. In this study, a conceptual framework is proposed to analyze qualitative feedback from students and classify them into 19 predefined aspects of Biggs’ model. Student feedback can be preprocessed using tokenization, stemming, and stopword removal. TextBlob was used to categorize the sentiment of students’ comments on each course using polarity and subjectivity. For the classification problem, a word embedding layer is used to transform the plain English words into vector representation and feed them to the deep learning model Bi-LSTM with forwarding and backward propagation. Deep learning is evaluated for its performance in multi-label classification. A case study with a desktop application adopting the proposed framework to analyze student comments of an education institution and illustrating the framework results in bar graphs. This would assist an educational institute in verifying its existing systems and improving its services to students. Overall, an application was designed for an educational institute to check and enhance teaching and learning practices.