position-bound and/or relation-bound? the productivity of nouns as
TRANSCRIPT
POSITION-BOUND AND/OR
RELATION-BOUND? THE
PRODUCTIVITY OF NOUNS AS
CONSTITUENTS OF COMPOUNDS.
Elizaveta Tarasova
Victoria University of Wellington
PhD Candidate
N+N COMPOUNDS
house mouse, thought police and library book
NOT hatchback
NOT singer-songwriter
WHY ARE COMPOUNDS INTERESTING FOR
LINGUISTS?
bear + paw = bear paw (bear HAS paws)
bear + scare = bear scare (? bear HAS scare)
citizen army vs. liberation army
security police vs. city police
shoe box vs. cardboard box
Baayen (2010):
the constituents of lexicalized compounds are
position-bound, i.e. they have a tendency for being
used either as a head or as a modifier more often.
Maguire, Wisniewski & Storms (2010):
nouns are used equally productively in the modifier
and head roles.
MAGUIRE, WISNIEWSKI & STORMS (2010)
RESEARCH QUESTIONS
The question about the morphological
productivity of constituents comprising a
compound
The question of whether the productivity of a
compound constituent on the morphological level
coincides with the productivity of the semantic
relation realized in the constituent family.
METHOD
100 nouns served as the basis for collecting a
corpus of constituent word families.
197 constituent word families comprising 7,332
compounds extracted from Davies (2004-) BNC.
The semantic relations between the elements of
the collected compounds analysed according to
Levi’s (1978) classification.
LEVI (1978) CLASSIFICATION OF RDPS
Meaning RDP Examples
N1 CAUSE N2 CAUSE1 Sex scandal
N2 CAUSE N1 CAUSE2 Tear gas
N1 HAVE N2 POSSESSION1 Lemon peel
N2 HAVE N1 POSSESSION2 Cameraphone
N1 MAKE N2 COMPOSITION1 Snowball
N2 MAKE N1 COMPOSITION2 Computer industry
N2 USE N1 INSTRUMENT2 Steam iron
N2 BE N1 ESSIVE2 Island state
N2 IS IN N1 LOCATION2 Field mouse
N2 IS FOR N1 PURPOSE2 Arms budget
N2 IS FROM N1 SOURCE1 Business profit
N2 IS ABOUT N1 TOPIC2 Tax law
HYPOTHESES
(1) A constituent of an N+N compound is used
more often in one position (modifier or head)
than the other.
(2) The constituent’s preference for one position
correlates with a tendency of its being used
in the same way semantically within the
constituent’s modifier or head paradigm, i.e.
there is a connection between the family size
and the constituent’s concentration on one
semantic relation.
DISTRIBUTION OF FAMILY SIZES
constituent minimum maximum mean standard
deviation
N1 0 83 31.79 20.47
N2 0 116 45.95 30.26
STATISTICS FOR HYPOTHESIS (1)
Model coefficients for linear model showing the
relationship between N1 and N2 family sizes.
N1 = 98, Adjusted R2 = 0.3796
Estimate Std. Error t value Pr(>|t|)
Intercept 3.43574 0.42932 8.003 2.65e-12 ***
logFamSizeN1 1.07457 0.32638 3.292 0.00139 **
(logFamSizeN1)2 -0.30143 0.06094 -4.946 3.17e-06 ***
RELATIONSHIP BETWEEN FAMILY SIZES
logFamSizeN1
logF
amS
izeN
2
1.5 2.0 2.5 3.0 3.5 4.0 4.5
2.5
3.0
3.5
4.0
4.5
HYPOTHESIS (2)
There is a connection between the family size and
the constituent’s concentration on one semantic
relation, i.e. the more productive a constituent is
in one position, the more likely it is to
demonstrate a preference for one semantic
relation.
The semantic content of a concept strongly
influences the way it is used in conceptual
combinations
e.g. ˂substance – artifact˃ → MADE OF
artifact MADE OF substance
box
frame
doll, etc.
plastic
wood
chocolate, etc.
SIMILAR MODIFIERS AND HEADS COMBINE
IN SIMILAR WAYS
Pairings of modifiers and heads are not
distributed randomly but fall into a number of
regular semantic patterns which reflect
productive semantic relationships
In the course of processing the meaning of a
compound, we utilize our inner statistical
knowledge, which is based on our previous
language experience
IMPLICATIONS FOR THE CURRENT
RESEARCH
The semantics of the concept can be an
influential factor for predicting the way the
constituent is likely to be used in a compound
If there is a connection between the semantics of
the noun concept and its productivity in one
position, then we should be able to see this
connection in the way this concept combines with
other concepts to form N+N combinations
The knowledge about how a given constituent
contributes to the overall meaning should be part
of the statistical knowledge
The semantics of the head noun serves as the
schema whose properties are responsible for
attracting certain modifiers based on the slots
that are more available for filling
For modifiers our previous linguistic and non-
linguistic experience of the modifying word is of
primary importance
HYPOTHESIS (2)
(2a) The constituent should demonstrate a
preference for certain relations, or even one
particular relation, over other relations.
(2b) This tendency should be stronger the larger
the constituent family is.
INSTANTIATIONS RATIO
The measure that is computed as the ratio of the
number of different semantic relations that a
given constituent realizes in its family and the
number of occurrences of the most frequent
relation in this family.
The smaller the Instantiations Ratio, the more
concentrated on one semantic relation the
compound constituent is.
heart + N compounds
future + N compounds
Diff. Relations Most Frequent Inst. Ratio
7 13 0.5384
Diff. Relations Most Frequent Inst. Ratio
4 77 0.051
STATISTICS FOR HYPOTHESIS (2)
Model coefficients for linear model showing the relationship between family size N1
and instantiations of relations ratio for N1. N1 = 98, Adjusted R2 = 0.1624
Model coefficients for linear model showing the relationship between family size N2
and instantiations of relations ratio for N2. N2 = 99, Adjusted R2 = 0.1934.
Estimate Std. Error t value Pr(>|t|)
Intercept 0.31094 0.13664 2.276 0.0251
logFamSizeN1 -0.18121 0.04071 -4.451 2.31e-05
Estimate Std. Error t value Pr(>|t|)
Intercept 0.21709 0.16684 1.301 0.196
logFamSizeN2 -0.22460 0.04538 -4.949 3.14e-06
WHICH CONSTITUENT IS MORE
IMPORTANT?
The constituent whose semantics is more
important for the meaning of the compound
should demonstrate consistency in realizing the
same relation in the constituent family and this
relation should be predictable from the semantic
content of the constituent concept.
E.G. COMMUNITY
“a group of individuals, plants and animals that live
in the same place or have a particular characteristic
in common” (OED)
Implied propositions
a community
consists of
members
who/what makes
the community?
(people, plants,
animals, etc.)
these
members live
in a certain
area/place
Where is the
community
located? (city,
village, area,
etc.)
The constituent that realizes this information
more consistently should be expected to be more
important for a compound to receive the reading
it receives
How to check that?
We can pick all N+N combinations where the
constituent is used and analyse the semantic
relations
INDIVIDUAL CONSTITUENT FAMILIES
The relation most frequently realized in head
families is always connected with the core
semantic content of the word.
The relational information does not seem to be so
dependent on the semantics of the modifier.
E.G. POLICE
the civil force of a state, responsible for the
prevention and detection of crime and the
maintenance of public order (Oxford online)
Key content – PURPOSE
or NEGATIVE PURPOSE
police + N – POSSESSION ×
e.g. police culture, police violence
N + police – (NEG) PURPOSE √ e.g. riot police, transport police
WHICH CONSTITUENT IS RESPONSIBLE FOR THE
RELATIONAL READING OF A COMPOUND?
The semantic relation consistently realized in head
families is directly connected with the semantic
content of the head noun
The semantics of the modifier is altered, the
semantics of the head is preserved
e.g. health centre community centre
centre (building)
WHICH CONSTITUENT IS RESPONSIBLE FOR THE
RELATIONAL READING OF A COMPOUND?
health service health food
health ?
(what we know about health)
SUMMARY
There is a connection between the productive use
of a compound constituent in one position and
productive use of the semantic relation, which
can be considered evidence of interaction between
the levels of structural and semantic
representation in compounds
There seems to be a direct connection between
the semantic content of the head noun and the
relation reading of a compound
The information that the modifier contributes to
the meaning of an N+N structure is peripheral to
the semantic content of the modifier concept
CONCLUSION
These findings may challenge the claim about the
leading role of the modifier in defining the
relational reading a compound receives and the
semantic weight of the modifier in an N+N
sequence.
Although there is no denying that the modifier
concept changes the meaning of the head concept,
the change of meaning of the semantic relation
seems to be connected with the semantics of the
head.
REFERENCES
Baayen, R.H. 2010. The directed compound graph of English. An exploration of lexical connectivity and its processing consequences. In S. Olsen (ed.), New impulses in word-formation (Linguistische Berichte Sonderheft 17), Hamburg: Helmut Buske Verlag. 383-402.
Davies, M. 2004-. BYU-BNC. (Based on the British National Corpus from Oxford University Press). Available online at http://corpus.byu.edu/bnc/.
Levi, J. 1978. The Syntax and Semantics of Complex Nominals. New York: Academic Press Inc.
Maguire, P., Wisniewski, E.J. & Storms, G. 2010. A corpus study of semantic patterns in compounding. Corpus Linguistics and Linguistic Theory 6(1), 49-73.