ENHANCING ENGLISH PRONUNCIATION AND LEARNER ATTITUDES VIA AIDRIVEN CAPT: AN EMPIRICAL STUDY ON VIETNAMESE SECONDARY SCHOOL STUDENTS

Abstract

This longitudinal mixed-methods study investigates the pedagogical efficacy and socio-affective outcomes of integrating an Artificial Intelligence-driven Computer-Assisted Pronunciation Training (CAPT) platform English Central within the secondary school English as a Foreign Language (EFL) context of Vietnam. The research explicitly addresses the systemic 'assessment paradox' where high-stakes national written examinations ignore formal oral components, resulting in fossilized phonological errors and intense language anxiety among learners. Utilizing a sample population of 30 eleventh-grade EFL students over an intensive eight-week intervention, quantitative speech analytics were captured via pre- and post-testing protocols using a validated 4-point analytical phonological rubric. This quantitative data was triangulated with multidimensional 5-point Likert scale attitudinal matrices. The experimental results demonstrate highly significant advancements across all evaluated phonological domains. Targeted first-language (L1) transfer barriers, specifically the substitution of interdental fricatives (/θ/, /ð/) and post-alveolar sibilants (/ʃ/), alongside the systemic omission of word-final consonant codas and complex clusters showed marked remediation (p < 0.01). Psychologically, the automated, non-judgmental Automatic Speech Recognition (ASR) feedback loop successfully lowered the participants' high baseline affective filter, protecting student selfesteem and fostering proactive autonomous practice habits outside school hours. The study provides actionable curricular frameworks for embedding adaptive language technologies into large-scale public school education to bridge the gap between error recognition and selfcorrection.