Large-scale characterization of Mandarin pronunciation errors made by native speakers of European languages

In this work, we quantify common tonal and phonetic errors made by second language learners of Mandarin Chinese. Pronunciation patterns of 300 native speakers of European languages are analyzed. Tonal errors (30.56%) are found to be more prevalent than phonetic ones (8.71%). Common errors include overemphasis of Tone 3 and inadequate aspiration of affricate consonants. Decision tree clustering was used to further characterize these error patterns with their tonal and phonetic context. Our findings are potentially useful in second language education and in computer assisted language learning.