Application of Machine Learning for Real-Time Feedback on Spoken English in Language Learning Apps
Abstract
Language learning apps have become integral tools for enhancing spoken English proficiency. Traditional methods of providing feedback in such platforms often rely on static, pre-programmed responses, which fail to adapt to individual learner needs. These methods lack the capacity for real-time, dynamic feedback that can address pronunciation, fluency, and context-specific language errors. This study proposes a novel approach by integrating GRU-GPT (Gated Recurrent Unit and Generative Pre-trained Transformer) models to provide real-time, personalized feedback for spoken English in language learning apps. The aim is to enhance the interactivity and responsiveness of learning platforms, providing users with accurate, context-sensitive corrections that cater to their learning pace and proficiency levels. The GRU model is used for its ability to capture temporal dependencies in speech, while GPT offers contextual understanding to generate meaningful corrections. The proposed method was chosen due to its superior ability to process sequential data (speech) and provide adaptive, meaningful corrections in real-time. The study's findings demonstrate that the GRU-GPT hybrid model outperforms traditional feedback systems by delivering more accurate, personalized corrections. Learners showed improved speaking fluency scores, pronunciation accuracy, and higher engagement levels, validating the efficacy of the proposed method. This work offers a valuable direction for more dynamic and efficient spoken English learning applications.