AVAR-RL: adaptive reinforcement learning approach for personalized English vocabulary acquisition

Abstract

The demand for personalized e-learning has surged, yet existing systems often fail to adapt to individual learners’ evolving needs. This paper introduces AVAR-RL, a novel reinforcement learning framework for English vocabulary acquisition that dynamically tailors learning paths using a Contextual Multi Armed Bandit approach. By integrating multi-dimensional learner profiles including proficiency, VARK styles, and real-time engagement AVAR-RL optimizes exercise recommendations in real time. Experiments with 600 ESL learners demonstrate 14.2% higher precision, 17.8% better retention, and 19.3% increased engagement compared to state-of-the-art baselines. The system’s scalability and cold-start performance (82.1% precision) make it a practical solution for adaptive e-learning platforms.