Identification of accent in spoken English by Cantonese ESL speakers

Abstract

Learners of English as a second language (ESL) often experience difficulty in correctly producing second language (L2) English sounds. Traditionally, accent in L2 sound production is detected by human perception—the ESL teacher or the learners themselves. This is apparently too subjective and can be unreliable for L2 learning. In this study, we made use of speaker-independent automatic speech recognition (ASR) system for objective accent identification in L2 English learning among Cantonese speakers. In the experiment, ESL speakers pronounced sonic English words and read out a short English passage. Patterns of error in L2 English pronunciation were identified by using an ASR software. Results reflected the error pattern in L2 English production and the strong influence of Cantonese as amother tongue among the speakers.