Young Korean EFL learners’ discrimination of AI-generated English front vowels

다국어 초록 (Multilingual Abstract)

This study evaluates the perceptual accuracy of young Korean EFL learners in distinguishing front vowel contrasts /i/, /ɪ/, /ɛ/, and /æ/ in the coda voicing contexts bVd and bVt. Utilizing advanced AI-generated speech stimuli from Amazon Polly, the study explores the discrimination challenges faced by these learners. Participants (N = 17) were subjected to an AB/BA discrimination task with a ternary forced-choice setup. Despite a high overall discrimination accuracy (mean = 87.01%), the learners found the /ɛ/-/æ/ contrast particularly more challenging than the /i/-/ɪ/ contrast. This challenge can be attributed to the closely matched F1 frequencies of AI-generated stimuli for /ɛ/ and /æ/.  Additionally, the inherent vowel length difference between /i/ and /ɪ/ likely served as a discernible cue, given its accessibility compared to spectral cues for young EFL learners. Results also indicate that participants more effectively discerned vowel contrasts following the voiced consonant /d/ than the voiceless /t/, suggesting that vowel acquisition in young Korean EFL learners can be enhanced in contexts with voiced codas, which extend vowel durations for better processing. The pedagogical implications of the study are that a targeted sequence of instruction, beginning with easier vowel contrasts like /i-ɪ/ in voiced coda contexts, can foster improved pronunciation. Additionally, incorporating AI tools and phonetic awareness exercises can further enhance English front vowel perception.