Sentiment analysis of students’ CQI: A comparative study using Textom and ChatGPT models (3.5 and 4.0)

Abstract

This study explores text-based analysis and advanced AI-driven sentiment analysis using GPT-3.5 and GPT-4.0 models to evaluate college students’ Continuous Quality Improvement (CQI) from students. The goal is to provide deeper insights into educational assessments by comparing and integrating both methods. Using Textom for keyword analysis, network visualization, and Ucinet6 NetDraw for CONCOR analysis, we processed a final dataset of 32,285 cleaned evaluations. Key terms such as "material," "test," "helpful," "liked," and "content" were identified through TF-IDF weighting, and the CONCOR analysis revealed one central opinion cluster and several sub-clusters focused on course content, teaching methods, and student participation. Additionally, sentiment analysis using GPT-3.5 and GPT-4.0 was conducted to categorize feedback into positive, negative, and neutral sentiments. The GPT-3.5 model demonstrated higher accuracy in understanding contextual nuances and detecting emotional intensity than traditional methods, highlighting areas of satisfaction like course materials and instructor engagement and identifying areas of dissatisfaction linked to evaluations and assignments. Integrating traditional Textom analysis and GPT-based sentiment analysis provides a comprehensive and actionable framework for understanding student feedback. This integration enables institutions to design targeted interventions, such as refining teaching practices, improving course content, and tailoring assessments to enhance student satisfaction and learning outcomes. The findings are particularly valuable in addressing challenges in remote and hybrid learning contexts, offering scalable solutions for adapting to evolving educational needs. By bridging traditional methods with AI-powered insights, this study underscores the transformative potential of AI in advancing academic quality.