Goodness of tone (GOT) for non-native Mandarin tone recognition

Lexical tone is one of the most challenging pronunciation problems in tonal language acquisition. Accurate lexical tone production is especially challenging for people whose native language is not a tonal one. In this paper, we propose Goodness of Tone (GOT), a confidence measure inspired from goodness of pronunciation (GOP) for tone recognition. GOT is a vector representation of the confidence of each lexical tone of the given speech segment. The proposed GOT confidence measure is useful in tone recognition due to the following: 1) Unlike other tonal features such as pitch or fundamental frequency variation, GOT integrates both phonetic and tonal information. 2) GOT exploits competing tonal phones which differ only in tonal label but are the same in phonetic labels as a reference to conduct cohort normalization. 3) GOT is a vector that concatenates confidence scores from all the possible lexical tones, making it easier to characterize error patterns of non-native tonal production.