Paper Contents
Abstract
AbstractLanguage diversity remains a major barrier to achieving equitable access to quality education, particularly for learners in low-resource language communities. Although digital learning platforms have broadened global access, most high-quality educational materials are still produced in a few dominant languages, limiting inclusion and learning outcomes. This paper proposes a Deep Learningbased AI framework for automated multilingual educational content generation. The system integrates Transformer-based Neural Machine Translation (NMT), readability-controlled text simplification, and neural Text-to-Speech (TTS) synthesis to produce context-preserving, semantically accurate, and linguistically accessible learning materials in multiple languages. A domain-adapted fine-tuning strategy is applied to ensure accurate translation of academic terminology, while data augmentation techniques improve performance for low-resource languages. The framework is validated through a case study using Class 910 Science content translated from English to Hindi, Bengali, and Tamil. Evaluation using BLEU, METEOR, COMET, and Mean Opinion Score (MOS), along with human feedback, demonstrates high translation fidelity, natural speech output, and improved learner accessibility. The results highlight the systems potential as a scalable solution for inclusive multilingual education and its relevance to global educational equity goals.Index Terms Multilingual Education, Neural Machine Translation, Deep Learning, Text Simplification, Text-to-Speech, Accessibility, Low-Resource Languages, Transformer Models, NLP, Educational Technology.
Copyright
Copyright © 2025 Kavya Singh . This is an open access article distributed under the Creative Commons Attribution License.
