Assistant Professor, Department of Psychology, Faculty of Psychology and Educational Sciences, Shahid Beheshti University, Tehran, Iran. , m-charifi@sbu.ac.ir
Abstract: (38 Views)
Background: Dyslexia is the most common type of specific learning disorder and can have adverse effects on students' academic performance and learning processes. Despite the limitations of conventional screening methods, studies involving voice analysis to assess reading skills in Iran are very limited.
Aims: The aim of this study was to design and evaluate a preliminary screening tool for reading disorder based on the analysis of Oral Reading Analysis using machine learning.
Methods: The research employed a descriptive-analytical method of the causal-comparative type. The statistical population of the study consisted of all students in Grades 1 to 3 of primary school in the cities of Tehran, Zahedan, Mashhad, and Qom during the 2024–2025 academic year. The sample group consisted of 586 students selected via convenience sampling from regular schools and comprehensive assessment, education, and rehabilitation centers in the aforementioned cities. In the present study, a separate text tailored to the specific grade level was prepared for each of the Grades 1 to 3 of primary school. Furthermore, differences in audio features between groups were analyzed using the Kruskal–Wallis testin SPSS software (version 27).
Results: The Kruskal–Wallis test revealed significant differences in all examined audio features among the normal, reading-difficulty, and dyslexic groups (P< 0.001). All four algorithms demonstrated acceptable performance in distinguishing the students, with overall model performance being higher in Grade 2. Logistic regression showed stable and desirable performance across all three grades and was selected as the final model, with AUC values of 0.96, 0.99, and 0.99 for Grades 1, 2, and 3, respectively.
Conclusion: Based on the findings of this study, If the data is expanded and enriched and multimodal models are utilized, this approach can serve as an effective tool for designing reliable, locally adapted intelligent systems for the early detection of dyslexia.
Article number: 1
Type of Study:
Research |
Subject:
Special Received: 2026/09/30 | Accepted: 2027/04/30
| Rights and permissions |
 |
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) License. |