Eng · 繁體 · 简体

News & Events

 News & Events Home
 News and Events Archive
Research Seminar: "Prosody in Text: Patterns of Word-length Choices in Written Chinese Corpora"
Jump To:
Topic:  Research Seminar: "Prosody in Text: Patterns of Word-length Choices in Written Chinese Corpora"
posted itemPosted - 18/02/2010 :  10:49:19


Department of Chinese, Translation and Linguistics

Research Seminar


Prosody in Text: Patterns of Word-length Choices in Written Chinese Corpora


Presented by


Professor San Duanmu

Department of Linguistics, University of Michigan




3 March 2010 (Wednesday)


4:30 - 6:00pm


P4704 (Level 4, Purple Zone), Academic Building, CityU






It is well known that Chinese has a preference for certain word length combinations, as shown in (1) and (2). The disfavored forms in [N N] and [V O] are the exact opposite. The patterns can be analyzed in terms of familiar metrical requirements, shown in (3)-(5).
 (1)    In [N N], 1+2 is not favored

         2+2       技術    工人        ‘skill worker’
         2+1       技術    工
       *1+2        技       工人
         1+1        技       工

 (2)    In [V O], 2+1 is not favored

        2+2         學習    繪畫         'study painting’
      *2+1         學習    畫
        1+2         學        繪畫
        1+1         學        畫

 (3)    Metrical requirements (S indicates a syllable)
a.   A foot is made of two syllables, either (S+S) or (SS).
b.  Phrasal stress goes to the first N in [N N] and O in [V O].

(4)    Foot structures for [N N] (offending foot underlined)
        (SS)(SS), (SS)S, *(S)(SS), (S+S)

(5)    Foot structures for [V O] (offending foot underlined)
        (SS)(SS), *(SS)(S), S(SS), (S+S)
Exceptions to (1) and (2) are possible (e.g. [皮 手套] 'leather glove' and [喜歡 錢] 'love money'. However, they are believed to be uncommon. The analysis makes a strong prediction that remains to be verified. In particular, we expect 1+2 to be infrequent for [N N] and 2+1 to be infrequent for [V O]. In this study I examine the prediction empirically, using the Lancaster Corpus of Mandarin Chinese, which contains a balanced variety of texts that total 1.5 million characters.

Initial statistics do not seem to support the prediction, but a closer examination shows that the prediction is largely correct. Quantitative results of all [N N] and [V N] patterns will be reported and analyzed, and exceptions will be discussed.




San DUANMU is Professor of Linguistics, University of Michigan. He received his Ph.D. in Linguistics from MIT in 1990 and has held teaching posts at Fudan University, Shanghai (1981-1986) and the University of Michigan (1991-present). He is the author of The Phonology of Standard Chinese (2nd edition, Oxford 2007) and Syllable Structure: The Liminits of Variation (Oxford 2009).


~ All Are Welcome ~



Enquiry: LTenquiry@cityu.edu.hk