An Entropy-based Study on the Acquisition of Mandarin Initial Consonants by Korean Learners

In the research field of second language (L2) speech acquisition, most studies use error rate or accuracy to measure learners' proficiency. However, such metrics cannot compare and uncover the error patterns of two phonemes with similar error rates. Using an entropy-based metric in [1], this study took Korean learners' acquisition of Mandarin initials as an example to examine the error distribution of L2 speech. We investigated the error dispersion of 52,576 Mandarin initials produced by 31 Korean Chinese learners (Three groups (G1, G2 and G3) were divided according to the duration of Chinese learning). The results showed that: (1) Among the 21 Mandarin initials with similar error rates, the one with a stable substitute from Korean had smaller error dispersion than the others. (2) With the increase of learning experience from G1 to G2, the decrease magnitude of error dispersion was greater than that of error rate. In conclusion, the entropy-based metric method can describe the dispersion of pronunciation errors influenced by the learners' first language and explain the process of Korean learners' Mandarin initial acquisition.