Optimizing English Language Evaluation: A Hybrid AI Framework Using Neural Networks and Fuzzy Decision-Making Models

Abstract

One of the most important aspects of educational evaluation is the accurate assessment of English language ability. This is especially crucial for ensuring that evaluations are fair and impartial. To improve the evaluation process of final scores for Iranian English as a Foreign Language (EFL) students enrolled in a reading comprehension course, this research proposes a hybrid artificial intelligence framework that incorporates both Neuro-Fuzzy Systems (NFS) and Artificial Neural Networks (ANN). The performance records of 66 students are included in the dataset, comprising scores for midterms, quizzes, finals, class participation, and bonus components. A neuro-fuzzy system, along with two-layer and three-layer ANN models, is used to train the hybrid framework to forecast students' final marks. These predictions are compared with scores assigned by instructors. Even though the three-layer ANN demonstrated higher accuracy than the two-layer version, the findings showed that NFS produced predictions most closely aligned with the aggregated instructor scores. These results highlight the potential of hybrid AI frameworks to enhance objectivity and reduce bias in academic evaluations. Overall, this study demonstrates that combining ANN with fuzzy decision-making models can improve intelligent scoring techniques, optimize evaluation procedures, and promote fairness in language assessment.