LFG and the Analysis of Chinese
organised by Adams Bodomo and Luke Kang Kwong
| This workshop aims at
investigating aspects of the structure of Chinese (including Mandarin and Cantonese) from
the perspective of the Lexical-Functional Grammar framework. With its unique properties,
such as having pro-drop but exhibiting little verbal morphology, being topic prominent
rather than subject prominent, permitting possibilities of long-distance binding, showing
considerable complexity in verb complementation, and more flexible word order than
English, etc., Chinese poses special challenges to linguistic description and theory. In
this workshop, these and related phenomena are discussed with a view to developing
innovative ways in which Chinese can be better handled using insights from LFG and other
constraint-based approaches. Implications for linguistic theory and language universals
will also be discussed.
Program of the workshop: Time scheduled: 14-17, June 27
The Subject Condition in Cantonese Luke Kang Kwong, Adams Bodomo, Owen
Nancarrow
The grammatical notion of subject occupies a key position in most linguistic theories. In configurational approaches, subject is defined either as [Spec IP] (or some other functional projections) (Haegeman, 1994) or generated in adjunction to VP (Stowell 1981, Chomsky 1986 and Manzini 1983). In LFG, subjects and other grammatical functions are regarded as primitives and represented in functional structures. The subject condition stipulates a default subject for every clause predicated by a finite verb. There are several statements of this condition in the literature. In Bresnan (2001:311) it is stated thus: “Every predicator must have a subject.” The status of the subject condition as a grammatical universal has been a matter of some controversy (see, for example, Alsina 1996, Bresnan and Zaenen 1990, Bresnan and Kanerva 1989, and Berman 1999). In this paper, we examine the subject condition with data from Cantonese. Cantonese, like all the other Chinese dialects, is a pro-drop language. However, unlike other pro-drop languages such as Italian, implicit subjects cannot be retrieved through verbal morphology, as shown in the following sentences.
In (1)-(4), it is not possible to determine from the verb forms the identity of the implicit subjects. In (1), the subject pronoun can be retrieved from the immediate speech context. In (2), the subject is understood to be the current topic of the conversation. (3), like all other meteorological sentences in the language, does not come with an expletive subject. Sentence (4) can be understood as either containing a zero-subject of generic reference (like ‘on’ in French) or having ‘sight-seeing’ as subject. There are also topic-comment sentences where the initial NP is not an argument of the main verb but is nevertheless related to it pragmatically or through the discourse context. The issue then is how to identify subjects in sentences where they do not have overt expressions. In this paper, we put forward a set of criteria with which some kinds of implicit subjects could legitimately be recovered. We propose that subjects should still be represented at f-structure in consonance with LFG approaches but that instead of achieving functional specification solely at the level of morphosyntax (constituent structure), subjects in Chinese should also be identifiable at a pragmatic-discourse level. This proposal is in line with Bresnan’s (2001:98) characterization of the subject as having “..the unique property of being both an argument function and a (grammaticalized) discourse function.”
References: Alsina, A. 1996. The Role of Argument Structure
in Grammar: Evidence from Romance. California: CSLI Publications.
An Optimality-Theoretic Account of Mandarin Complex Reflexive ‘ta-ziji’ (s/he-self) Haihua Pan, Jianhua Hu
As noted in Pan (1995, 1998), non-contrastive complex reflexive ta-ziji in Chinese (i) can have a long-distance (LD) bound antecedent (cf. 7); (ii) allows non-c-command/sub-command antecedents (cf. 10); and (iii) also observes some kind of blocking effect (cf. 8-9). Although Pan (1998) gives a correct description of the binding properties of ta-ziji, he does not explain why Bill in (7) does not block the binding of ta-ziji to John, though Bill does not appear lower than John in his Animacy Hierarchy. In this paper we will show that the binding properties exhibited by ta-ziji can be best explained if we adopt an Optimality-theoretic account of reflexivization. We think that the blocking effect of ta-ziji can be derived from the prominence constraint which stipulates that the most prominent NP will block the binding of reflexives. Under our analysis, the interpretation of ta-ziji is regulated by the ranking (cf. 5) of the following constraints:
An anaphor cannot be coindexed with an NP that dominates it. Two coindexed elements must have compatible features. An anaphor cannot be bound to a across b which is not less prominent than a . Select the closest NP as the antecedent of an anaphor. I-within-I Constraint (IC) >> Feature Compatibility Constraint (FCC) >> Prominence Constraint (PC) >> Locality Constraint (LC) We think that the prominence of a NP is determined by two factors: [± Subject], [± Human]. The interaction of these two factors will yield the following results of which (6a) is the feature specification of the most prominent NP. b. [-SUBJ, +HUM] The constraint ranking given in (5) can correctly predict the binding possibilities of ta-ziji in the following sentences: say DE little-trick hurt-Perf he-self think I like he-self think I like he-self
The coindexations between the reflexive and the embedded subject NP in (7) and (8) are ruled out by FCC. Although the coindexation between Bill and the reflexive in (7) violates LC, it does not violate any other higher-ranked constraints, and is thus a better candidate than the embedded subject which violates FCC. The coindexation between John and the reflexive in (9) is also ruled out because it violates LC and PC, whereas Bill violates none of the constraints. The coindexation of John with the reflexive in (10) satisfies the higher ranked constraint FCC, and hence it is better than the subject which violates FCC. Note that our Optimality-theoretic account can explain why the coindexation between John and the reflexive is blocked in (8-9) but not in (7), since the intervening NP in (8-9) is an NP which is not less prominent than John, but none of the intervening NPs in (7) is such an NP, according to (6).
LFG for Chinese: Issues of Representation and Computation Sun Maosong LFG is quite powerful in describing linguistic phenomena of Chinese. Even a relatively sophisticated sentential construction could be successfully explained by LFG. Take sentence 1 as an example:
On the other hand, I show that LFG, as a computational formalism, is still not strong enough for computing Chinese. The fact that Chinese is an inflection-free language (for instance, neither a change in form nor an explicit marker is used when a verb functions as the main verb, clause, infinitive, modifier of nouns, or head of noun phrases) may result in a large number of ambiguities at every linguistic level for computers. The mapping between c-structure and f-structure, as well as the mapping between f-structure and a-structure are extremely difficult to build up, if semantic information is not provided sufficiently. Consider a group of sentences: 2a. Zhang-san da3 si3 le gou3. person drink drunk AUX wine N1 V1 V2 AUX N2 Zhang-san drank (the wine), and (Zhang-san) got drunk. 2c. Zhang-san ku1 zhong3 le yan3jing1. The c-structures of these three sentences are patterned in the same way, but their f-structures are quite different: in (2a), N2 is both OBJECT of V1 and SUBJECT of XCOMP; in (2b), N2 is still OBJECT of V1, but SUBJECT of XCOMP becomes N1; in (2c), N2 serves only as SUBJECT of XCOMP, no longer OBJECT of V1. It is nothing but semantic constraints, among V1, V2, N1 and N2, that control the one-to-many mapping processes from c-structure to f-structure. Similar cases are frequently encountered in Chinese. Consider another group of sentences which concerns “V1+N1+de+N2”, a popular syntactically ambiguous construction in Chinese: 3a. Yao3 lie4ren2 de gou3 3b. Yao3 lie4ren2 de ji1 3c. Yao3 tu4zi3 de gou3 The situation can be even complicated if word segmentation ambiguities are included. Incorrect segmentation may still lead to a syntactically well-formed but semantically ill-formed ‘sentence’. The point addressed here is that semantic analysis is likely to be in a dominant position in computing Chinese sentences. Manipulations on a-structure, f-structure and c-structure should be carried out jointly and in parallel. To render LFG truly computable for Chinese, I believe that some augmentation is needed accordingly:
References [1] Joan Bresnan. 2001. Lexical-functional syntax.
Oxford: Blackwell Publishers.
On the function COMP in Cantonese Adams Bodomo and Sophia Lee The existence of COMP as a grammatical function within LFG has been the subject of much discussion. Alsina, Mohanan and Mohanan (1996a, b & c) propose that the clausal complements, which are commonly considered as bearing COMP, can be taken as bearing the OBJ function. Other works consider that there is evidence for COMP as a distinct group (L2 drup 1996; Curly 1996; Dalrymple and L2 drup 2000). In this paper we propose to contribute to this discussion with analysis of data on Cantonese COMPs. We first point out the possible problems for Alsina et al.'s proposal for eliminating COMP. These include the existence of non-configurational languages, the status of XCOMP, and the inability of COMP to passivize in Cantonese. Following Dalrymple and L2 drup (2000), we next take up the concept of a mixed language using Cantonese data. Initial results show that Cantonese is indeed a mixed language. We do however propose that the concept of mixed languages should be extended to include degrees or extents to which a language can be regarded as a mixed language. Our analysis in this paper supports a finer-grained categorization of grammatical functions in linguistic theory.
About the speakers Adams Bodomo is Assistant Professor at the Department of Linguistics, University of Hong Kong. He specializes in theoretical linguistics, particularly syntactic theory, lexical semantics, and the syntax-semantics interface. He has done research on complex predicates, serial verb constructions, and other verbal phenomena across various languages, including Dagaare, Twi, Cantonese, French, and Norwegian. Kang-Kwong Luke is Senior Lecturer and Head, Department of Linguistics, University of Hong Kong. His areas of research include Phonetics and Phonology, Gammar and Lexis, Sociolinguistics and Conversation Analysis. He has worked on English, Cantonese, and other Chinese dialects. Owen Nancarrow is Lecturer in the Department of Linguistics, University of Hong Kong. His expertise is in Phonology, Morphology and Syntax. He has worked on English, German, Cantonese and Bantu. Haihua Pan is Assistant Professor at the Department of Chinese, Translation and Linguistics, City University of Hong Kong. He specializes in theoretical linguistics, specifically syntactic theory, formal semantics, and computational linguistics. He has done research on a variety of topics including focus and negation, reflexive binding, aspect, argument structure, the passive construction, topicalization, noun phrase extraction, subject identification, etc. Jianhua Hu is a Ph D candidate at the Department of Chinese, Translation and Linguistics, City University of Hong Kong. He specializes in theoretical linguistics. He has done research on topics such as anaphoric binding, topicalization, and pronoun resolution. Sun Maosong is Associate Professor and Associate Head of Department of Computer Science, Tsinghua University (Beijing). His research interests include computational linguistics, statistical and corpus-based natural language processing, Chinese computing (morphology, syntax and semantics), information retrieval and machine translation.Sophia Lee is currently an M.Phil. student in the Department of Linguistics at the University of Hong Kong. She works on a comparative study of Cantonese and English syntax within the theoretical framework of Lexical-Functional Grammar [Back to Abstracts index] |