<![CDATA[This study investigates how Artificial Intelligence (AI) systems recognize the English interdental fricatives /θ/ and /ð/ in Indonesian-accented speech. Because these sounds are absent from the Indonesian phonological inventory, Indonesian learners often experience difficulty producing them, which may affect intelligibility and AI recognition. Using a qualitative phonetic analysis with AI-based comparison, speech data from six Indonesian learners of English and one native speaker were collected. The recordings were analyzed using Praat to examine acoustic characteristics and OpenL to generate speech-to-IPA transcriptions. The results show that many learner productions lacked sustained fricative turbulence, indicating non-target realizations of interdental fricatives. OpenL generally reflected these acoustic deviations rather than producing canonical forms, suggesting limited sensitivity to subtle fricative cues. Overall, the findings reveal an intersection between human phonetic challenges and technological limitations in current ASR systems, highlighting the need for accent-inclusive training data and focused pronunciation instruction to improve both intelligibility and AI speech recognition performance. Keywords : Interdental fricatives; Indonesian-accented English; OpenL]]>
Copyrights © 2026
