Centroid-Based Aspect Sentiment Analysis for MOOC Learner Reviews
Type
Supervisor
Subject
Date
2025-05-01
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
Developing comprehensive analytics for Massive Open Online Courses (MOOCs) is essential for improving course design and enhancing learner engagement. In this work, we introduce MOOCSense, a multi-stage sentiment analysis module designed to analyze MOOC learner reviews and contribute to generating detailed MOOC analytics. In the first stage, we employ a mapping algorithm that extracts key MOOC-specific terms and central semantic phrases from the reviews. In the second stage, we propose a novel Centroid-Based Learning approach combined with the BERT (CLB) model to capture both implicit and explicit sentiment polarity in learner reviews, leveraging BERT’s deep contextual understanding of natural language. By focusing on the central semantics of each review, our approach uncovers the emotional drivers behind learner engagement or dissatisfaction. This dual-stage module enables more accurate sentiment association with specific course aspects, enriching MOOC analytics with valuable insights. Experimental results demonstrate the effectiveness of our approach across various MOOC datasets, achieving an accuracy of 92%, making it a promising solution for generating in-depth learning analytics and supporting course improvement strategies.