Pronunciation errors of Korean nasal codas by Japanese learners: A statistical study using AI training data for foreign voice recognition
Kyung Min Moon, Soo-Yeon Kim · The Journal of Linguistics Science · 2023
This study analyzes AI training data from Japanese learners regarding their pronunciation errors of Korean nasal codas. Two types of speech data sets were examined: reading speech and spontaneous speech. Approximately 38 hours of Korean speech data from 40 Japanese speakers indicated that Japanese learners of Korean exhibited a high error rate in pronouncing nasal codas due to L1 negative transfer effects. Additionally, this study investigated whether orthographic factors play a role when Japanese learners of Korean pronounce consecutively positioned nasal codas. It was found that assimilation error rates in reading, when consecutive codas were presented, were higher than those observed in spontaneous speech.