TR-VABML: Enhancing Turkish vocabulary acquisition through adaptive machine learning classification

Abstract

Conventional vocabulary assessments emphasize precision rather than hesitation and rapidity. A machine learning system was developed utilizing behavioral analysis and linguistic insights to identify vocabulary gaps in Turkish language learners. This system integrates hesitation counts, reaction times, and answer attempts with word difficulty and thematic elements. Vocabulary strength was computed using a rule-based equation derived from behavioral indications. With 89% accuracy, 86% precision, 91% recall, and an 88% F1 score, the model showed better performance than the linear and Poisson kernel alternatives. By effectively separating complex interactions, the RBF kernel minimizes unnecessary actions and ensures accurate identification of real shortages.