AI AUGMENTED NEURAL FEEDBACK FOR PRECISION ACQUISITION OF NON-NATIVE PHONOLOGY IN SYNCHRONOUS DIGITAL LEARNING ENVIRONMENTS

The ongoing challenge of learning non-native phonology in the synchronous digital learning setting is a primary obstacle to intelligible pronunciation and communicative competence. Conventional online teaching is rarely accompanied by real-time, personalized corrective feedback, leading to the persistence of fossilized pronunciation errors and poor phonological accuracy. The proposed study is an Artificial Intelligence (AI)-augmented neural feedback model that will be used to improve the accuracy of the acquisition of non-native phonological features by leveraging adaptive acoustic modelling and real-time articulatory feedback. The system combines deep neural networks trained on 12,500 annotated speech samples across 8 phonological categories and allows detection of segmental and suprasegmental deviations, achieving a mean phoneme recognition accuracy of 94.3%. In a synchronous virtual classroom environment, 120 learners were split into 60 control and 60 experimental groups during an 8-week intervention period, which was used to conduct a quasi-experimental evaluation. Students who received AI-enhanced neural feedback showed 31.8% higher phoneme-level accuracy, 24.6% lower prosodic deviation scores, and a statistically significant increase in intelligibility ratings (p < 0.01) compared with students receiving traditional instructor-based feedback. Sustained, real-time interaction was achieved with a latency of under 180 ms. Additionally, scores for learner engagement improved by 22.4%, demonstrating greater motivation and ongoing participation. Data-driven neural feedback systems coupled with AI and synchronous frameworks greatly improve learning accuracy for learners' phonology and put a promising approach to non-native pronunciation systems within the context of intelligent language learning.

Hali tarjima qilinmagan