ClusSumm a user centered hierarchical clustering and adaptive transformer framework for efficient summarization of MOOC learner reviews

Abstract

Massive Open Online Courses (MOOCs) have become a major source of online education, generating large volumes of learner reviews that provide valuable insights into course quality and learning experiences. However, manually analyzing this feedback is time-consuming, while existing transformer-based summarization models often struggle to balance summary conciseness with information preservation. In addition, most evaluation frameworks rely solely on automatic metrics and rarely consider user preferences. To address these limitations, this study proposes ClusSumm, a user-centered hierarchical clustering and adaptive transformer framework for summarizing MOOC learner reviews. A domain-stratified dataset containing approximately 8,000 Coursera course records across eight educational domains was constructed through automated web scraping. Eight state-of-the-art transformer models, including BART, PEGASUS, T5, FLAN-T5, DistilBART, LED, LongT5, and BigBird PEGASUS, were benchmarked to establish comparative baselines. ClusSumm first applies K-Means clustering with silhouette-based optimization to group semantically similar reviews, followed by TF-IDF sentence ranking to remove redundant content. An adaptive summarization strategy then assigns clusters to different transformer models according to their complexity, and a hierarchical BART-based fusion module generates the final summary. Experimental results show that ClusSumm achieves a summarization ratio of 12.65% while maintaining ROUGE-1 and ROUGE-2 precision scores of 98.31% and 90.12%, respectively. A user-calibrated composite evaluation further demonstrates that the proposed framework produces concise summaries that better align with user preferences than existing benchmark models. These findings highlight the effectiveness of combining semantic clustering, adaptive summarization, and user-centered evaluation for scalable educational feedback analysis and intelligent learning analytics.

The authors' abstract, as published at the source. Discover Artificial Intelligence, 2026 · DOI ↗