Dual-System Recommendation Architecture for Adaptive Reading Intervention Platform Tailored for Dyslexic Learners

Abstract

Dyslexia poses substantial literacy challenges with profound academic and psychosocial impacts for affected children. Though evidence affirms that early reading interventions can significantly improve outcomes, traditional one-size-fits-all approaches often fail to address students’ unique skill gaps. This study details an adaptive reading platform that customizes word recognition tasks to each learner’s evolving abilities using embedded recommender engines. Initial standardized assessments categorize words by difficulty and cluster students by competency level. An integrated word generator then expands the benchmark lexicon by algorithmically manipulating phonetic properties to modulate complexity. Dual intra-user and inter-user systems track learner performance to tailor content to individuals’ pacing. Heuristic bootstrapping and simulated user data facilitate cold start recommendations and evaluate model robustness. Analysis of five virtual student response patterns demonstrates platform reliability against volatility. Successive interventions display narrowing score dispersion alongside upwards literacy trajectories. Logarithmic score pro-gressions signify responsive tuning to emerging mastery, accelerating ad-vancement, and tapering gains as maximal outcomes reached. Results validate system effectiveness in optimizing challenge levels to unlock growth for neuro-logical diversity. Rapid stabilization around optimal zones signifies an efficiently learned model while improved achievement confirms scaffolding precision. Learning curves substantiate tailored recommendation efficacy and signal user transitions from constructing new knowledge to demonstrative skill gains. Overall, the approach shows immense promise in administering personalized, engagement-focused reading support.