Boosting EFL Learners' Speaking Skills and Motivation Through Dual Graph Neural Networks

Effective communication and the capacity to express oneself precisely are central to English as a Foreign Language (EFL) courses. Successful language acquisition is now often measured by the ability to communicate effectively orally. Previous studies have shown that insufficient teaching methods and a lack of motivation are the main causes of EFL students' low speaking proficiency. There is limited opportunity for teachers to innovate or offer fresh viewpoints to their teachings due to the course’s standardized curriculum and the traditional, inflexible, and tedious teaching methods employed in speaking classes. As a result, pupils lack interest in the classes. Included in the suggested method are the steps of preprocessing, feature extraction, and model training. The normalization of the signal should precede the addition of frames to the speech signal during preprocessing. Mel-Frequency Cepstral Coefficients (MFCCs) were employed for feature extraction. All through the course, STDGNN was always in use. This novel approach outperforms CNN and GNN with an average accuracy of 91.63 percent.