9th Annual Conference of the International Speech Communication Association

Brisbane, Australia
September 22-26, 2008

A Comparative Study on Dissyllabic Stress Patterns of Mandarin and Cantonese

Weixiang Hu (1), Jin Jian (2), Aijun Li (3), Xia Wang (1)

(1) Nokia Research Center, China; (2) Sun Yat-sen University, China; (3) Chinese Academy of Social Sciences, China

This paper studied the similarity and dissimilarity for stress patterns between standard and various dialects of Mandarin, focusing on pitch and duration pattern comparison. We analyzed the distribution, pitch and duration patterns for stressed dissyllabic words in Cantonese, Cantonese-spoken Mandarin and Standard Mandarin. For isolated dissyllabic words, it was shown that there was a preference for a stress on the preceding word in Standard Mandarin while there was a stress on the posterior word in Cantonese-spoken Mandarin, and it is more easily lengthened in Cantonese-spoken Mandarin than in Standard Mandarin. For dissyllabic words in an utterance, there was a preference for stress on the preceding word in both Standard and Cantonese-spoken Mandarin but it is more easily lengthened in Standard Mandarin. Such results are due to the impact of the Cantonese dialect. At the same time, we investigated the neutral tone realization in Cantonese spoken Mandarin utterances. It was found that 36.6% of the neutral tones were wrongly realized in the other tones. It is an important reason for the difference of in stress patterns in Standard Mandarin and Cantonese-spoken Mandarin. It is necessary to compare acoustic cues according to the stress level. It is not enough for determining the similarities and dissimilarities for the rhythmic pattern between two dialects or between Mandarin and accented Mandarin according to the basic two stress pattern (stress-unstressed and unstressed-stressed.)

Full Paper

Bibliographic reference.  Hu, Weixiang / Jian, Jin / Li, Aijun / Wang, Xia (2008): "A comparative study on dissyllabic stress patterns of Mandarin and Cantonese", In INTERSPEECH-2008, 1105-1108.