Posters
Author: Agata Guskaroska (Iowa State University)
The purpose of this study is to examine Automated Speech Recognition (ASR) software and its potential for facilitating vowel pronunciation practice for Macedonian English as a Foreign Language (EFL) learners. A list of 12 sentences including minimal pairs of the contrasts /i/-/ɪ/, /æ/-/ɛ/, /u/-/ʊ/, and /ɑ/-/ʌ/ was recorded by 10 Macedonian learners, aged 18-19 and two American English native speakers in order to test the reliability of ASR. The speech samples were turned into text using ASR and the results of the written output were compared between native speakers and non-native speakers. Results demonstrated that the program was accurate in transcribing most of the vowel sounds for native speech. ASR written output was less accurate for non-native speech and was most likely indicating learners’ mispronunciations of vowels by transcribing them inaccurately. The results suggest that ASR may be promising for individual vowel practice but future research may involve words in isolation to avoid the system’s flaws in making assumptions based on context.
Keywords:
How to Cite: Guskaroska, A. (2018) “The Potential of ASR for Facilitating Vowel Pronunciation Practice for Macedonian Learners”, Pronunciation in Second Language Learning and Teaching Proceedings. 10(1).