A study of relations between associative structure and morphological
structure of Hungarian words
Dániel Czégel Zsolt Lengyel
Budapest University of Technology Pannon University, Veszprém
and Economics
czegel d@yahoo.com
Csaba Pléh
Central European University, Budapest
vispleh@ceu.hu
The paper mainly aims to reanalyze data with do
put
hammer
sit
down
the presently available corpus linguistics tools old
sit
go
quick
slow square
from a relatively large scale paper-and-pencil newspaper
catch
sight of holdcome
walk
clock
chair
based Hungarian verbal association dictionary
keep hour
send go do enter
see watch page
run
away lesson side
purchase to bed
money daughter paper
with regard to two aspects. i) The mental lexicon
work stand difficult town
corner
earnlook forsearch find heavy baby
male woman doll
man footleg short stem
sound little street
hand
issue. How are associative overlaps representing
teach think voice soldier strong high
job people village
say tall
can girl lion ADJ
know to name
look at
ADJMODN
structural relations in the mental lexicon? ii) The
young power place long
speak to space ADJN
learn help group part big mountain
understand answer acquaintance promise free square house
boygrandmother great ADVN
to ask known simple turn
window river
love allowed country
man
familiar INF
systemic variability of the associative fields mo-
like child outdoors
word excuse father table
to hungarian
give have to head
beautiful
roompicture
carpet MOD
milk new
must
hear book
interesting
shape deep MODV
tranquil
need guest smooth
bilized by the stimulus words: how variable the
life
to bad dear forest N
world road
write
russian kind impression matter
law friend good dear way eye ocean
NV
earth
result time
expensive
thief mother order
clean V
responses are, and how these associative entropies
mark to weather
truth school
law flower
ticket
remember joy green
wish
to besilent holiday
wrath healthdream year
to listen to doctor yellow blue
dot
are related to morphological entropies of the same
loud physician to moon
music
live point colour
problem dark
score white lamp
hungry red black
day
words. thirsty
to
sleep bread
production
soft
hard
light
sun
stove
drink salt fruit
sweet water
war
to morning full
eat sour cold
illness bitter
1 Methods and materials
cinema
butter
stomach
saturday
For the associative corpora, two dictionaries of Figure 1: Associative field of children (age 10-14)
Lengyel [3] were used. They are based on the
responses of 2000 students between 10 − 14 and
18 − 24 to about 200 stimulus words. Digitized square
responses from this dictionary were related to the hammer
quick part
frequency distribution of 800 million web-based
go chair
paper
side
year
law run clock way
thief street
enterhour
Hungarian words from the MOKK corpus [2].
country
production lesson music
hungarian square stem
sit down
find do full short
daughter
baby space
sit stop city
foot doll head
little
heavy
hold name mountain
word milk
come send high ocean
law hand
to picture table
cinema
school village
write catch group long house
carpet window ADJ
2 Results
do
sight of world
time big river ADJ MOD N
weather boy room
voice
say order place ADJ N
know book new sound
think smooth wood Adv N
work man figure
answer learn resultwork
teach matter interestingsimple free yellow
russian beautiful
young eye flower colour INF
promise to child girl allowed walk earth
speak ask help man outdoors blue MOD
to lamp
good
2.1 Associative overlaps and lexical fields
ticket
understand to be truthto clean green red moon N
look for wanting lifewoman
to live joy corner NV
to like dream sun
give hear father bad
acquaintance friend
to
look at to dear day
dark
V
remember thanquil strong black lion
to listen to stand dear
guest health white
to be silent bed light see
excuse to mother
Based on the associative overlap measure intro-
news wish impression hard
sleep
money stove
newspaper feast bread
buy morning water
saturday power
to trouble
deep
duced by Deese [1], a multidimensional scaling
eat soldier
go away sweet
salt slow
wrath cold
grandmother soft
people loud
method was used to obtain associative fields de-
butter full
drink hungrybitter fruit
thirsty
doctor
picting the pairwise associative distance of stim- war
illness
sour stomach
old
ulus words in a two-dimensional figure. The re-
sults indicate that young adults have a more dense
Figure 2: Associative field of young adults (age
structure, their associative clusters are more tight
18-24)
compared to those of children of age 10-14, as il-
lustrated in figure (1) and (2) and shown quantita-
tively in figure (3) and (4).
Copyright c by the paper’s authors. Copying permitted for private and academic purposes.
In Vito Pirrelli, Claudia Marzi, Marcello Ferro (eds.): Word Structure and Word Usage. Proceedings of the NetWordS Final
Conference, Pisa, March 30-April 1, 2015, published at http://ceur-ws.org
117
sociative entropy and the logarithm of corpus fre-
quency: In the case of nouns the correlation is pos-
1000
itive (r = 0.134 and 0.367 in the two ages) while
frequency
the correlation is negative (r = −0.281, −0.222)
100 for verbs. This peculiar relation would be further
studied with considering morphological entropy in
10 light of the argument frames of the verbs on the
one hand, and the role of syntagmatic associations
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14
in the associative fields of verbs on the other [4].
associative overlap 7
æ
6 æ æ ADJ
Figure 3: Histogram of pairwise associative over-
morphological entropy
æ
æ
5 æ æ æ æ 8 ADJMODN
æ ææ
æ
laps (age 10-14)
æ
ææ æ
ææ æ æ ææ
ææ æ
æ
æ
æ æ ADJN
æ ææ ææ æ
4 æ æ æ æ æ æ æ æ
ææ
æ
æ
æ æ æ ææ
æ
æ
æ æ æ æ æ
8 AdvN
æ æææææ æ
æ æ æ æ æ
æ ææ æ ææ
3 æ æ
æ
æ
æ æ æææ æ æ æ æ æ
æ INF
æ ææ ææ ææ æ
æ ææ æ æ æ
10 000 2
æ æ æ
æ æ
æ
æ ææ
æ æ ææææææ
æ æ æ
æææ
æ
ææ
æ æ ææ 8 MOD
æ ææ æ æ æ æ
æ æ æ æ æ æ æ æ N
æ æ æ ææ æ æ æ
æ ææ
æ
1 æ ææ
ææ æ æ
æ ææ
ææ æ
æ
æ 8 NV
æ æ æ
æ æ
1000 0 æ æ æ
æ æ
æ æ V
3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0
frequency
associative entropy
100
Figure 5: Relation between associative and mor-
10 phological entropy (age 18-24). The blue regres-
sion line corresponds to nouns, while the red line
corresponds to verbs.
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14
associative overlap
Figure 4: Histogram of pairwise associative over- 106 æ
æ æ
æ
ADJ
laps (age 18-24)
æ
æ ADJMODN
æ
æ æ æ
æ ææ æ ææ ææ æ ADJN
frequency
æ æ æ æ æ
105 æ
æ æ æ æ
ææ æ
æ ææ æ æ
æ æ
æ æ ææææææ
æ
æ æ æ ææ æ æ æ
æ ææ
æ AdvN
æææ æ æ ææ ææ æ æ æ ææ æ ææ æ æ
æ æ ææ æ æ
æ æ æææ ææ æ æ
æ
ææ æ ææ æ æ
æ æ
ææ ææææ æ
æ æ
æ æ æ
æ
æ
æ æ INF
æ
ææ æ æ æ
2.2 Relationship between associative entropy,
æ ææ
104 æ
æ
æ æ ææ ææ
æ
æ æ æææææ ææ
æææ
æ
æ ææ
æ
æ
MOD
æ
æ æ æ ææ æ æ
æ æ æ
N
morphological entropy and frequency æ æ ææ æ ææ æ
æ æ
æ
æ æ ææ
1000 æ NV
V
We followed the methods introduced by Osgood
3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0
associative entropy
[5] for analyzing the variability of morphologi-
cal and associative structure of words. As shown Figure 6: Relation between associative entropy
in figure (5), there is an interesting difference of and corpus frequency (age 18-24). The blue re-
the relations between associative entropy (defined gression line corresponds to nouns, while the red
P
as HA = − i pi log2 pi , where pi is the rel- line corresponds to verbs.
ative frequency of the ith associated word) and
corpus morphological entropy (defined as HM =
P
− i qi log2 qi , where qi is the relative frequency References
of form i in the MOKK corpus) between nouns
[1] James Deese. The structure of associations in
and verbs. In nouns, the more varied the morphol-
language and thought. Johns Hopkins Univer-
ogy of the noun is in the corpus, the more variable
sity Press, 1966.
the associative field is (r = 0.202, 0.175 in the
two ages). That can be interpreted as implying that [2] Péter Halácsy, András Kornai, László
the more varied the suffixation of a noun is, the Németh, András Rung, István Szakadát,
more variable associative relations it enters with and Viktor Trón. Creating open language
other words. In verbs, however, if the verb has a resources for hungarian. In LREC, 2004.
more varied morphology, it has less associations [3] Zsolt Lengyel. Magyar asszociációs normák
(r = −0.194, both groups). As figure (6) shows, a enciklopédiája I-II. TINTA könyvkiadó, Bu-
similar relationship has been obtained between as- dapest, 2008-2010.
118
[4] Katherine Nelson. The syntagmatic-
paradigmatic shift revisited: a review of
research and theory. Psychological Bulletin,
84(1):93, 1977.
[5] Charles Egerton Osgood. Cross-cultural uni-
versals of affective meaning. University of Illi-
nois Press, 1975.
119