Learning to Spell Smarter: A Deep Q-Learning Framework for Personalized Spelling Education

Abstract

Digital spelling applications require learners to practice words from lists that stay constant regardless of whether the user is excelling or falling behind. This research implements Adaptive Artificial Intelligence (AI) to create a spelling bee training application that continuously adjusts based on user input to keep users learning at the optimal level of difficulty. This application powered by Deep Q-Learning (DQL), adjusts the difficulty of words given to users based on measurable performance metrics such as accuracy, speed, and number of retries. Technologies used to build the application include React, Flask, MySQL, and Generative AI (used to supply definitions, usage, semantic understanding, and audio pronunciation). Neuroplasticity concepts are implemented by having the reinforcement learning agent continually provide learners with content within their zone of proximal development. Better word retention occurred because the learner had to think about the word more in order to understand and remember it.

Keywords: Spelling education, Reinforcement learning, Adaptive Learning, Deep Q-learning, Learning Analytics, Personalized Learning