What question did this study set out to answer?

The research aims to develop a resource-aware framework to enhance personalization in English language teaching.

February 11, 2026Open Access

A Resource-Aware Multi-Agent Reinforcement Learning Framework for Personalized English Teaching

LSLinzhi ShaoChengdu University of Technology

Key Points

The research aims to develop a resource-aware framework to enhance personalization in English language teaching.
Developed a MultiAgent Proximal tuned deep edge Q-learning (R-MAP-DEQL) framework.
Used datasets including vocabulary, grammar exercises, and pronunciation audio samples.
Implemented data preprocessing techniques like tokenization and normalization.
Extracted Mel-Frequency Cepstral Coefficients from audio for better speech pattern analysis.
Utilized Python with reinforcement learning and deep learning libraries for implementation.
Achieved 97% accuracy, 95% precision, 93% recall, and 96% F1-score.
Experimental results showed improved performance across vocabulary, grammar, and reading metrics.
Demonstrated enhanced student engagement and personalized lesson adaptations.
Visualizations confirmed the system's effectiveness in improving language skills.

Abstract

English language teaching (ELT) systems often lack personalization and adaptive feedback. Traditional English teaching methods lack personalization, real-time feedback and engagement. Integrating Multi-Agent Reinforcement Learning (MARL) offers adaptive, interactive learning environments. Agents include Teacher, Student, Content, Evaluation and Interaction modules, collaboratively learning optimal teaching strategies. Lack of reward-based personalization limits adaptive lesson selection, reduces engagement and weakens real-time feedback, hindering effective English learning. Research aims to develop a resource-aware MultiAgent Proximal tuned deep edge Q-learning (R-MAP-DEQL) framework for a personalized English teaching system. Datasets include English vocabulary, grammar exercises, reading passages and audio samples of pronunciation. Data preprocessing involves tokenization and normalization of text to standardize input and remove noise. MelFrequency Cepstral Coefficients (MFCC) are extracted from audio samples to capture pronunciation and speech patterns. DEQL with proximal tuning enables agents to optimize policies efficiently, balancing exploration and exploitation while accounting for computational constraints and providing real-time personalized teaching interventions. The framework is implemented in Python using RL and deep learning libraries. Experiments demonstrate improved learner performance, engagement and personalized lesson adaptation. Visualizations show progressive improvement across vocabulary, grammar, reading and pronunciation metrics, confirming system effectiveness. Experimental results demonstrate that R-MAP-DEQL achieves an accuracy of 97%, a precision of 95%, a recall of 93%, and an F1-score of 96%. The proposed MARL-based English teaching system effectively personalizes learning, adapts dynamically to student performance and enhances engagement. Resource-aware multi-agent (MA) strategies ensure optimized teaching decisions. Results highlight potential for AI-driven education, scalable real-time deployment and improved language skill acquisition.

Ask AI

Helpful

Bookmark

View Full Paper