A Decade of Adaptive Learning in Online Education: A Text Mining-Based Systematic Mapping

Abstract

Modern education requires curriculum flexibility to support lifelong learning. Differences in learners’ backgrounds, learning styles, and goals have increased the need for adaptive learning systems that provide personalized, responsive learning experiences. Advances in machine learning, natural language processing, and predictive learning models have further accelerated the development of adaptive learning environments. This study maps the development and research trends of adaptive learning in online education from 2014 to 2024 using a Latent Dirichlet Allocation (LDA)-based text-mining approach. We quantitatively and thematically analyzed 887 Scopus-indexed publications. The results show a consistent increase in publication volume, with China and the United States emerging as the most productive contributing countries. Model comparison using topic coherence and log perplexity supported selecting a ten-topic LDA model. The analysis identified artificial intelligence, virtual platforms, and learning outcomes as the most dominant themes. Other major themes included learner characteristics and cognitive modeling, personalized pedagogical design, mobile learning tools, recommendation systems, learning analytics, digital assessment, teacher feedback, and professional development. Semantic mapping showed that personalized pedagogical design occupied an integrative position across technological, pedagogical, and learner-centered research strands. Temporal topic evolution further revealed a shift from early learner- and system-oriented discussions toward emerging concerns related to digital assessment, cognitive skills, teacher development, and technology-supported learning practices. These findings provide a structured overview of the intellectual landscape of adaptive learning in online education and identify directions for future research.