Verb compounds within canonical typology: Chinese separable verb compounds

1
Verb compounds within canonical typology
Chinese separable verb compounds

• Anna Siewierska
• Jiajin Xu
• Richard Xiao

2
Overview of the talk
1
Separable verb compounds (SVCs)
2
Canonical typological strategy
3
A case study of SVCs in Mandarin
2
3
Separable verb compounds

• Some languages have verb compounds which are made
  up of two parts, a verbal stem and a movable
  element standing before or after the verb in
  adjacency or close proximity
• Different terms in the literature
• separable verb compounds, split words, separable
  verbs, ionised words, discontinuous / detachable
  / breakable / discrete words, etc

3
3
4
An example of Chinese SVC

• dan1xin1, lit. carry heart, to worry
• dan1-le yi1 shang4wu3 xin1, carry ASP one morning
  heart, to be worried the whole morning
• xin1 yi4zhi2 dan1-zhe, heart all the time carry
  ASP, to have been worried all the time

4
5
Sound similar?

• Derivation by infixing (e.g. abso-fucking-lutely)
  and syntactic interposing (e.g. of bloody course)
  in English
• Separable complex verbs in Dutch (aankomen
  arrive) and German (ankommen arrive)
• But Chinese SVCs are

5
5
6
essentially different

• 1) Insertions in English infixing and interposing
• Almost exclusively restricted to expletives,
  euphemisms, and amplifiers
• Acting as an emotive intensifier
• In contrast, discontinuous use of Chinese SVCs
  has a greater variety of insertions and discourse
  / pragmatic functions
• Insertions as head / tail satellites aspect
  markers, RVCs, quantifiers, classifiers,
  modifiers, etc
• Providing extra information
• Acting as a mitigator / softener
• Showing casualness
• Expressing negative emotions such as disapproval
• Enhancing rhythm important in a syllable-timed
  language like Chinese

6
7
essentially different

• 2) A significant difference between SVCs in
  Mandarin and the split prefix phenomenon in Dutch
  (e.g. binnenkomen, to come in) and German (e.g.
  abfahren, to drive off/depart)
• Chinese SVCs are not words with a separable affix
• E.g. dan1xin1 worry
• V O

7
8
essentially different

• 3) SVCs in Dutch and German can have a wide range
  of constituents of all types as insertions,
  including complex NGs and subordinate clauses as
  in the example below
• A Dutch example of opbellen ring up
• Ik bel op
• Ik bel hem op
• I ring him up
• Ik bel hem morgen op
• I ring him tomorrow up
• Ik bel de man waarvan ik houd op
• I ring the man that I love up
• ...which is completely impossible in Chinese

8
8
9
Why are SVCs interesting?

• 1) SVCs are a large class of verbs in Chinese
  which cannot be marginalised
• 2) They satisfy none of the universal criteria
  for wordhood (Dixon and Aikhenvald 2002 19-20)
• A grammatical word consists of a number of
  grammatical elements which (a) always occur
  together, rather than scattered through the
  clause (the criterion of cohesiveness) (b) occur
  in a fixed order (c) have a conventionalised
  coherence and meaning
• Criterion (c) means that speakers of the language
  may talk about a word (but are unlikely to talk
  about a morpheme)

9
10
Why are SVCs interesting?

• 3) SVCs violate one of the most fundamental
  principle of the theory of word formation
• The Principle of Lexical Integrity Word-internal
  structures are not accessible to rules of syntax
  (Booij 1990 45)
• 4) SVCs are listed as words, but they clearly
  have some phrasal properties, thus straddling
  the boundary of morphology and syntax
• E.g. the analysable internal structures of
  Chinese SVCs

10
11
Canonical typology

• To study such fuzzy and cross-border grammatical
  categories, canonical typology (CT) has proved to
  be a useful strategy (cf. Bond 2007 Corbett
  2007 Nikolaeva 2008), e.g.
• Suppletive forms
• Agreement
• Negation
• Syncretism

11
12
Standard strategy in typological research (Croft
2003 14)

• Determine the particular structure or situation
  type of interest
• Examine the morpho-syntactic construction(s) or
  strategies used to encode that situation type
• Search for dependencies between the constructions
  used for that situation and other linguistic
  factors
• i.e. other structural features and external
  functions expressed by the structure, or both

12
13
Canonical typological approach

1. Start with a linguistic phenomenon
2. Establish a general definition for identifying
   that linguistic category
3. Construct a set of features or criteria for the
   typical (canonical) case of the category
4. Use the criteria to investigate the relevant
   categories in languages

13
14
Canonical typological approach

1. Start with a linguistic phenomenon
2. Establish a general definition for identifying
   the linguistic category in question
3. Construct a set of features or criteria for the
   canonical case of the category
4. Use the criteria to investigate the relevant
   categories in languages

14
15
How can corpora inform CT?

• In CT, the features are usually collected from
  the literature
• The collection could be selective, subjective and
  arbitrary
• Can the selection of features be more objective
  and reliable?
• We seek to answer this question from the corpus
  linguistic perspective
• The corpus-based approach makes it possible for
  variational parameters of SVCs to be summarised
  exhaustively and more objectively by looking at a
  large amount of attested language use
  simultaneously

15
16
A case study of Chinese SVCs

• What are common types of insertions and external
  patterns of discontinuous use of SVCs in
  Mandarin?
• How can canonical features be identified on the
  basis of frequency?
• How can the study of SVCs in Chinese contribute
  to the research of similar phenomena in other
  languages?

16
17
Prevalence of SVCs in Mandarin

• The 2002 edition of the Modern Chinese Dictionary
  includes 3,236 types of SVCs (Zhu 2006 29)
• Four categories verb-object (97),
  verb-complement, subject- predicate, and
  coordinative
• Given their prevalence, no grammar of Chinese can
  turn a blind eye to the verb-object paradox
  (Packard 2003 108)

17
18
Corpora

• Two corpora are used in this study
• The Lancaster Corpus of Mandarin Chinese (LCMC)
  for written Chinese
• The Lancaster Los Angeles Corpus of Spoken
  Chinese (LLSCC) for spoken Chinese
• The LCMC is a balanced corpus of written Chinese
  composed of one million words proportionally
  sampled from fifteen genres ranging from news,
  fiction to academic prose published in mainland
  China around 1991 (see McEnery, Xiao Mo 2003)

18
19
Corpora

• The LLSCC comprises one million words of
  dialogues (55) and monologues (45) in Chinese,
  covering both spontaneous (57) and scripted
  (43) speech in six spoken genres
• The two corpora are also tokenised and POS-tagged
• They provide an empirical basis for our
  quantitative and qualitative analysis of SVCs in
  Chinese

19
20
Seed SVCs for data extraction

• A total of 1,738 commonly used SVCs listed in A
  Dictionary of Split Word Usage in Modern Chinese
  (Yang 1995) were used as seeds to automatically
  extract all instances of possible SVCs
  exhaustively when their the head and tail are
  separated, in either forward or backward
  direction, by a span of 1-10 words
• 2793 raw concordance lines were extracted from
  the two corpora

20
21
Human evaluation and annotation

• Each concordance line was evaluated independently
  by two native Chinese speakers in order to remove
  noise in automatically extracted results
• Only 565 true instances of discontinuous use of
  SVCs are retained for further annotation and
  analysis
• Type of insertion, direction of separation, word
  semantics, sentence semantics (i.e. pragmatic
  meaning), sentence type, genre

21
22
Syntagmatic pattern of SVCs
22
23
Head satellites of SVCs

• Aspect insertion
• Expanded aspect insertion
• Note The ? slot can be filled or left blank

Pattern SVC types () SVC tokens ()
SVCH-le SVCT 42 (25) 74 (13)
SVCH-guo SVCT 15 (9) 22 (4)
SVCH-zhe SVCT 12 (7) 35 (6)
Total 69 (42) 131 (23)
Pattern SVC types() SVC tokens()
SVCH (?) ASP (?) SVCT 91 (55) 244 (43)
23
24
Head satellites of SVCs

• RVC insertion
• Expanded RVC insertion
• hardly surprising given that RVCs can be
  analysed as markers of the completive aspect in
  Chinese (Xiao and McEnery 2004)

Pattern SVC type () SVC token ()
SVCH RVC SVCT 20 (12) 26 (5)
Pattern SVC types () SVC tokens ()
SVCH (?) RVC (?) SVCT 20 (12) 66 (12)
24
25
Tail satellites of SVCs

• Classifier (CL)
• 21 (116 SVCs) contain a classifier
• Nominals in Mandarin are typically preceded by a
  classifier
• Quantifier (MC)
• 19 (108 SVCs) contain a quantifying construction
• Modifier (MOD), i.e. pre-modifiers of tails
• Possessive pronouns (64 times, 11)
• Adjectival modifiers (63 times, 11)
• Nominal items (59 times, 10)
• Question word (i.e. shen2me what, 26 times, 5)
• Also combinations of these elements

25
26
SVC networkLexical and grammatical patterning
26
27
Words or phrases?

• Synchronically, located somewhere on the
  continuum between words and phrases (cf. Guo and
  Qian 2004)
• words SVCs idioms phrases
• Diachronically, wordhood subject to language
  change
• Many compound words in current use have evolved
  from phrases (e.g. daoqian apologise, jugong
  bow)
• Givón (1971) Today's morphology is yesterday's
  syntax.
• Two criteria - depending on the type and number
  morpheme(s) in the insertion
• Over half of discontinuous use of SVCs in our
  data (i.e. 54 if RVCs are seen as quasi-aspect
  markers), together with their combined cognates,
  can be analysed as legitimate compound words

27
28
Two overarching criteria

• Structural criteria
• Host dependency
• Head dependence enjoys priority over tail
  dependence
• Phonological criteria
• PrWd restriction (Feng 2001, 2002)
• A disyllabic unit is the typical prosodic foot in
  Chinese
• A trisyllabic unit can also be a prosodic word

28
29
Structural criteria

• According to the host dependency criteria of the
  canonical typological approach
• a) SVCs with a clitic-like aspect marker alone
  are compounds rather than phrases
• b) SVCs with an RVC attached to the head verb as
  quasi-compounds
• c) Other modifiers (classifiers, modifiers, etc)
  attached to the tail (represented typically by a
  object or complement) are least possible
  compounds
• Priority a gt b gt c

29
30
Phonological criteria

• Various manifestations of SVCs define a continuum
  of phonological conditions which complement the
  structural criteria
• a) The combined uses of head and tail are
  disyllabic compounds
• b) SVCs in which the head and tail are separated
  by one single morpheme are possible compounds
  under the Trisyllabic Foot Rule (TFR) of prosodic
  morphology (McCarthy Prince 1993 1995)
• c) The head and tail separated by polymorphemic
  insertions like quantifiers, adjectival modifiers
  etc are phrases
• Priority a gt b gt c

30
31
Conclusions

• We have used the corpus-based approach to
  generalise canonical internal structures of
  Chinese SVCs
• The structural and phonological criteria we have
  proposed work well to define wordhood of SVCs in
  Mandarin
• The approach combining canonical typology and
  corpus methodology could also be useful in
  research of similar phenomena in other languages

31
32

• Thank you!
• Richard.Xiao_at_edgehill.ac.uk