Presentation on theme: "On the Nature of Word Classes in Chinese K.K. Luke Nanyang Technological University."— Presentation transcript:
On the Nature of Word Classes in Chinese K.K. Luke Nanyang Technological University
Debates on word classes Started in the 1950s – Gao Mingkai, Lü Shuxiang and others on the relative importance of morphological and syntactic criteria Latest one in 2009 (Yuyanxue Luncong) – Shen Jiaxuan, Zhan Weidong and others A span of more than 50 years
The first debate As a result of the first debate, it was generally agreed that the main criterion for identifying part-of-speech should be syntactic function, not morphology. Gabelentz (1881): The existence of grammatical categories is proved by the fact that Chinese words differ in their syntactic behaviour. But few have dared to take this argument to its logical conclusion.
Reluctance to follow through Famous example of chuban publish/ publication publish books the publication of books a ceremony to announce the publication of a book, i.e., a book launch Reluctance to assign chuban to two or more classes ( )
More recent debates Everybody agrees that the purpose of setting up word classes is for the convenience of doing grammatical analysis, but have word classes brought us more convenience or headache? (Preface to the 2009 special issue of Yuyanxue Luncong) Contributions informed by Linguistic Typology and Natural Language Processing (NLP)
Word classes for NLP A degree of arbitrariness in assigning POS tags No two corpora use the same POS tagset – PKU corpus: 38 tags – Academia Sinica Balanced Corpus: 43 tags – Penn Tree Bank: 33 tags No guarantee (or hope) that the POS tags used in NLP will correspond in any way to how words are represented and organised in speakers minds.
Sources of problem Why cant scholars agree on word classes in Chinese? – Structure of the language – Psychology of the linguist
Structure of the language Little inflectional morphology – Word classes have little morphological marking Flexibility – chi: chifan eat (rice); xiaochi snack – zao: da zao early morning; Zao! Morning!; zao shuo Why didnt you say that earlier? – Jun jun, chen chen, fu fu, zi zi (Confucian Analects) –
Psychology of the linguist Fear: that Chinese might be regarded as an inferior language without grammar if it turns out that words cannot be assigned to fixed word classes. ( ) W. von Humboldt (1826): The Chinese language seems rather to disdain than to neglect the denoting of grammatical categories.
Psychology of the linguist Great reluctance to entertain the possibility that syntactic categories can be determined only by reference to constructions. Li Jinxi 1950 ( ) (Words do not belong to any syntactic classes until they enter into a construction.) Li Jinxis view has almost been universally rejected for fear that it may lead to an unwanted conclusion: Words dont belong to unique word classes in Chinese.
Unnecessary worries But: – First, no fixed word classes doesnt mean no word classes; and – Second, no word classes doesnt mean no grammar.
One word, one class? Great effort made to ensure one word, one class and rule out the possibility of class overlapping (e.g., Lu Jianming 1994). – When all members of a class can occur in a non- typical position (e.g. All verbs can take subject or complement position) laodong work: laodong guangrong, xihuan laodong – Temporary shifts Tai junfa le! Thats too warlord-like! – Different meanings suo to lock/ a lock; daibiao; baogao
Noun or Verb? Possible cases of overlapping – Yanjiu research, diaocha survey, chuban publish Morphology no help – chuban to publish/publication – yanjiu to research/a piece of research – Nothing in the form of these words tells us whether they are nouns or verbs. Nominalization? – General reluctance to treat chuban in zhe ben shu de chuban as overlapping or nominalization
Diaocha in PKU corpus (v) to survey languages (vn) language survey (vn) conduct a survey (v/vn) large-scale survey (v) through surveying X/ through a survey of X
Different views Huang Changning – Class overlapping Guo Rui – Verbs (priority treatment – youxian kaolü) Reasons: economy, psychological acceptance Shen Jiaxuan – Verbs as nouns
Huang Changning X-bar theory: Head of an NP should be an N No good reason why words like chuban and yanjiu should not be treated as nouns, just like any other word in the same syntactic position: – zhe ben shu de chuban – [ NP X de N]
Guo Rui Words like chuban and diaocha are verbs. Distinction between lexical meaning and syntactic function Unlike other languages, in Chinese verbs (and adjectives) can simply occupy the Head position of a NP without undergoing any licensing process, e.g., nominalization. Overlapping is acceptable only if its rare.
Shen Jiaxuan A V N Noun as a superclass Relationship between N and V: constitution as opposed to realization (as in derivational morphology) E.g., English realise (v) > realisation (n); cf. Chinese shixian (v/n)
A constructionist approach Im more in sympathy with Huang, i.e., N/V overlapping However, I would add that words belong to different classes by virtue of their occurrence in different constructions. This idea is adopted from William Crofts Radical Construction Grammar (RCG) As Randy LaPolla has pointed out, RCG can be used to good effect in analysing Sino-Tibetan and Austronesian languages.
Radical Construction Grammar Key reference: Croft (2001) RCG RCG is a nonreductionist theory which begins with the larger units and defines the smaller ones in terms of their relation to the larger units. (2001: 47) Constructions, not categories and relations, are the basic primitive units of syntactic representation. (2001: 46)
RCGs conception of language The proper definition of speech community is a population of individual speakers who are communicatively isolated from other speakers. The communicative interaction of speakers defines another population: the population of utterances produced by the speakers in a speech community. A language is a population of utterances – not possible utterances, but actual utterances, just as the species is a population of actual organisms. (2001: 365)
POS in RCG Noun and verb are not universal categories. Word class labels are a convenient way of referring to classes of words in a particular language. Nouns and verbs of different languages could (and usually do) have very different properties, e.g., – Have a look/smell (English) – Hen junfa (Chinese), cf. ?very warlord
Constructions Form-meaning pairing Form: syntactic elements Meaning: semantic components Link between form and meaning: symbolic The role of each element determined with reference to the construction
Example Wangmian 7-years-old at died father Wangmians father died when he was seven The LOSS construction: X LOST Y Verbs that can take second position come from a small collection: At the age of seven, Wangmian lost his father
Solution Role of chuban determined by place in a particular construction – publish books (VO) – publication of books (X de Y) – book launch (NN) If the same word can occur in different constructions, then it will have different syntactic roles defined by those constructions. Original question whats its word class? is a misleading question.
Unnecessary worries Scanty inflectional morphology – Syntactic position more important Flexibility – Languages probably all flexible each in its own ways No word classes does not make Chinese an inferior language
Flexibility English, for example, has a great deal of overlapping too – Most simple nouns can be used as verbs Chair, table, butter, name, home, light, house, eye, plant – Most simple verbs can be used as nouns (in the Have a X construction) Look, walk, say, think, try
Conclusion Old paradigm – Words can be assigned to a small number of word classes according to their syntactic behaviour, and they are stored in the mental lexicon with word class tags attached to them. New paradigm – Words have meaning potential. By virtue of their meaning potential they can enter into particular positions in particular constructions and play roles defined by those constructions.