Student Feedback Sentiment Analysis and Classification Using the Enhanced Model

Abstract

Over recent years, the advancement of sentiment analysis has significantly influenced numerous fields, including business, social networking, and education. In the educational context, its growing adoption reflects an increasing interest in understanding and interpreting students’ opinions and feedback. Nevertheless, processing such information presents notable challenges due to the informal nature of student language and the vast quantity of data produced in learning environments. Sentiment analysis uses machine learning algorithms to classify text or data by detecting underlying attitudes and emotions, allowing for the detection of positive, negative and neutral sentiments within text-based collections. In this work, we hope to create an improved model capable of properly detecting the underlying sentiments in students' responses. This method allows educators to assess student satisfaction with the learning resources and teaching activities of the Flipped Blended Teaching (FBT) model. To increase overall student satisfaction with the new FBT model, the proposed system incorporates sentiment classification techniques from machine learning and natural language processing. Closed-ended questions were used to elicit feedback from teachers on their satisfaction with the FBT model. After implementation, the proposed sentiment analysis model, which uses the Multilayer Perceptron (MLP) algorithm, obtained a prediction accuracy of 98.4 %, exceeding all other algorithms evaluated.