Educational and e-learning environments must incorporate Spoken Language Understanding (SLU) and Automatic Speech Recognition (ASR) to ensure accessible, scalable, and inclusive digital learning. Existing algorithms struggle with low-resource languages, accents, and linguistic variety, making them unsuitable for real-world multilingual educational environments. Therefore, this study proposes EduLinguaNet, a Multilingual Conformer Network for online classes and schools, which solves these issues in accurate speech detection. The proposed EduLinguaNet employs an end-to-end Conformer Neural network architecture for precise translation and semantic comprehension across multiple languages, utilizing layer-by-layer neural networks for local feature modeling and self-attention for long-distance interactions. The proposed model is trained and assessed utilizing large-scale multilingual educational corpora, language learning platforms, interactive e-learning sessions, and lecture recordings. Experimental data show that EduLinguaNet outperforms Transformers and Recurrent Neural Networks (RNNs) in intent recognition and Word Error Rate (WER). Robustness in low-resource language settings and code-switching performance indicate the system’s adaptability for diverse learners. Finally, EduLinguaNet improves e-learning via accurate, real-time, multilingual speech detection and interpretation.
AlMashhadani et al. (Thu,) studied this question.