Reinforcement Learning in Personalized Vocabulary Training for Esl Students
Abstract
Individual vocabulary training is very important as it helps improve English as a Second Language (ESL) by modifying the acquisition of words individually to suit the needs of the learner. Reinforcement Learning (RL) is an efficient model to maximize this personalization through the continuous adaptation of learning probes according to the performance of the learner. Nevertheless, the current methods of vocabulary training typically use frozen lists and either repetitive drilling methods, which do not consider the diversity of learners, their engagement and long term memory. In order to address the limitations, the developed approach uses an advanced RL algorithm, Proximal Policy Optimization (PPO) to dynamically balance exploration and exploitation when choosing vocabulary tasks. In this framework, it is possible to present the words in an adaptive way, practice by spacing, and practice based on the context, which guarantees the effective and learner-specific acquisition of vocabulary. The given strategy has the benefit of enhancing retention as well as encouraging motivation and using contextual words. The results of the experiment show that mastering vocabulary is better, the level of conscious participation of learners is better, and the results of the training process are better than when using traditional training methods.