Evaluating the Effectiveness of Speech-to-Text Technology in Improving Pronunciation among ESL Learners

Abstract

This study examines how well Speech-to-Text (STT) Technology actually works as a tool for helping ESL learners improve their pronunciation. Despite the fact that speech recognition apps are now everywhere in education, we still have solid empirical evidence about whether they genuinely help learners produce more accurate speech sounds. Using a Quasi-experimental design, this study tracked pronunciation gains over a twelve-week period, comparing learners who used STT-mediated feedback against those in traditional classroom settings. The investigation looked at multiple dimensions of pronunciation: segmental accuracy, suprasegmental features and overall intelligibility, using both established rating scales and acoustic analysis. The results showed that the STT group improved significantly more in segmental accuracy and developed stronger self-monitoring habits, though gains in suprasegmental features were noticeably smaller. When asked about their experience, learners consistently mentioned that they valued the immediate, non-judgmental feedback the technology provided, even while acknowledging its limitations with prosodic nuance. These findings point toward a complementary role for STT tools rather than using them as a standalone solution. The study adds to the growing CAPT literature by providing empirical backing for weaving STT applications into ESL curricula.