The impact of AI-driven speech recognition on listening comprehension and pronunciation accuracy in English language teaching

This study aims to assess the impact of AI-driven information technology on English language acquisition relative to traditional methods. Fifty first-year students (mean age = 28; 38% male, 62% female) at Osh State University in Kyrgyzstan were randomly assigned to an experimental group (AI-assisted learning, n = 25) or a control group (conventional teaching, n = 25). The Vosk Automatic Speech Recognition (ASR) technology was employed to monitor learner performance over 12 weeks, utilising Oxford vocabulary sets. The results indicate that the experimental group demonstrated significant improvements in their listening skills. For instance, there was a 16.6% increase in word recognition accuracy (p = 0.003), a 2.7-point improvement in sentence comprehension (on a 10-point scale) (p = 0.012), and a 16.5% reduction in pronunciation error rates (p = 0.004). Nonparametric ANCOVA revealed that the IT group significantly outperformed the control group (mean score = 6.45 ± 1.12 vs. 4.18 ± 1.05, p < 0.05), demonstrating a small to medium effect size (η² = 0.09). Both techniques demonstrated effectiveness (p < 0.05 for within-group gains), with AI-based learning yielding statistically significant advantages. The findings align with previous research regarding the potential of educational technology to create more learner-centred environments, demonstrating that information technology can address shortcomings in conventional teaching approaches. The study’s limitations include a restricted sample size and an emphasis on listening abilities; subsequent research should investigate broader language competencies. This study emphasises the transformative potential of AI in language education and advocates its integration into curricula to improve learning outcomes.