The Headedness of Mandarin Chinese Serial Verb Constructions: A Corpus-Based Study

Copyright 2012 by Jingxia Lin, Chu-Ren Huang, Huarui Zhang, and Hongzhi Xu 26th Pacific Asia Conference on Language,Information and Computation pages 428–435 The Headedness of Mandarin Chinese Serial Verb Constructions: A Corpus-Based Study Jingxia Lin1 Chu-Ren Huang1 Huarui Zhang1, 2 Hongzhi Xu1 1The Hong Kong Polytechnic University; 2Peking University {ctjlin; churen.huang}@polyu.edu.hk; hrzhang@pku.edu.cn; hongz.xu@gmail.com Abstract Existing treebanks of Mandarin Chinese such as the Sinica Treebank, the Harbin Institute of Technology Treebank, and the Penn Chinese Treebank, parse Chinese serial verb constructions incorrectly or inconsistently in terms of headedness, i.e. which verb to be assigned with the label of syntactic and/or semantic “head”. Aspectual markers in serial verb constructions can help determine the head of these constructions (Li, 1991; among others). However, the majority of Chinese serial verb constructions do not have overt aspectual markers. Based on large-scale corpus studies, this work investigates the distribution of aspectual markers in Chinese serial verb constructions in order to explore which verb in the serial verbs is more likely to function as the head, and thus provides a reference for parsing serial verb constructions without overt aspectual markers. We find that contrary to previous studies such as Collins (1997), Law and Veenstra (1992) and Sebba (1987) that treat the first verb in a serial verb construction as the head, Chinese serial verb constructions more often have the second verb as the head. The results of this work can not only serve as a reference for automatic parsing of Chinese data, but also shed light on theoretical studies of the structure of serial verb constructions in Chinese and other serial verb languages. 1 Introduction This section first introduces the kind of serial verb constructions (SVCs) under discussion; then, it shows the difficulties in identifying the head verb of an SVC, both in terms of theoretical linguistics and automatic parsing. 1.1 The SVCs of this study “Serial verb construction” is not a unified notion in previous studies (Sebba, 1987; Lord, 1993; Durie, 1997; Aikhenvald, 2006; Li and Thompson, 1981; Paul, 2008; among many others). This paper focuses on SVCs in Mandarin Chinese, in particular, the type of SVCs in a narrower scope and is usually treated as typical SVCs by previous studies. Such SVCs display the following properties (cf. Li and Thompson, 1981; Paul, 2008; Muller and Lipenkova, 2009; Zhang, 2010; among others):1 (i) An SVC consists of a sequence of VPs with no overt connective markers; these VPs express simultaneous or immediately consecutive actions that can be conceived as one event. (ii) The VPs share the same grammatical subject. For instance, (1a) is an SVC, but (1b) is not because the object of the first verb (V1) is the subject of the second verb (V2). (1) a. chuqu kai-men exit open-door ‘go out to open the door’ b. qing ta he-cha invite him drink-tea ‘invite him to drink the tea’ (iii) The VPs in an SVC can occur as the main VP in a clause. For instance, (2a) is an 1 The majority of Chinese SVCs …

Content truncated to 500 words