study of relations between associative structure and structure of Hungarian words morphological

Abstract

mountain ocean table carpet window river ADJ ADJ MOD N ADJ N ADV N INF MOD MOD V N N V V Copyrig Claudia

Central

Budapest

The paper mainly aims to reanalyze data with the presently available corpus linguistics tools from a relatively large scale paper-and-pencil based Hungarian verbal association dictionary with regard to two aspects. i) The mental lexicon issue. How are associative overlaps representing structural relations in the mental lexicon? ii) The systemic variability of the associative fields mobilized by the stimulus words: how variable the responses are, and how these associative entropies are related to morphological entropies of the same words.

materials For the associative corpora, two dictionaries of 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 responses from this dictionary were related to the frequency distribution of 800 million web-based Hungarian words from the MOKK corpus [ 2 ]. 2.1

Results

Associative overlaps and le xic al fields Based on the associative overlap measure introduced by Deese [ 1 ], a multidimensional scaling method was used to obtain associative fields depicting the pairwise associative distance of stimulus words in a two-dimensional figure. The results indicate that young adults have a more dense structure, their associative clusters are more tight compared to those of children of age 10-14, as illustrated in figure (1) and (2) and shown quantitatively in figure (3) and (4). do put old hammer

sit catch sight of go do away money work sound voice go sit down quick slow walk clock hour watchbed les on hand girl power place big great new worldroad order way year thirsty drink to eat sweet sour to sleep il nes stomach salt bit er war brefarudit production soft

morning cold but er cinema saturday

square chairpage side paper lit floeot leg high tal

town corner short stemstreet vil age lion table rowomipnisdcsqptotuuuawracrenree calropnegt house countrymountain smooethye forest deep ocean earth clean

yel ow rwedhimtebolaocnkdark

light hard water ful flower green bluecolour day sun stove lamp river

Figure

1:

Associative field of children (age 10-14) look for law do know answer square thirsftyruit doctor war il nes sour stomach old wood lion colour lamp stove stem 2:

Associative

field young adults

(age In

Vito ht c by the paper’s authors. Copying permitted for private and academic Marzi, Marcello Ferro (eds.): Word Structure and Word Usage. Proceedin Conference, Pisa, March 30-April 1, 2015, published at http://ceur-ws.org 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14

associative overlap

Relationship between associative entropy, morphological entropy and frequency We followed the methods introduced by Osgood [ 5 ] for analyzing the variability of morphological and associative structure of words. As shown in figure (5), there is an interesting difference of the relations between associative entropy (defined as HA = − Pi pi log2 pi, where pi is the relative frequency of the ith associated word) and corpus morphological entropy (defined as HM = − Pi qi log2 qi, where qi is the relative frequency of form i in the MOKK corpus) between nouns and verbs. In nouns, the more varied the morphology of the noun is in the corpus, the more variable the associative field is (r = 0.202, 0.175 in the two ages). That can be interpreted as implying that the more varied the suffixation of a noun is, the more variable associative relations it enters with other words. In verbs, however, if the verb has a more varied morphology, it has less associations (r = −0.194, both groups). As figure (6) shows, a similar relationship has been obtained between associative entropy and the logarithm of corpus frequency: In the case of nouns the correlation is positive (r = 0.134 and 0.367 in the two ages) while the correlation is negative (r = −0.281, −0.222) for verbs. This peculiar relation would be further studied with considering morphological entropy in light of the argument frames of the verbs on the one hand, and the role of syntagmatic associations in the associative fields of verbs on the other [ 4 ]. 7 6 y trop 5 n l 4 e a c i g looh 3ae ro 2 p m1 ae 106 ae aeae aeaeaeaeaeae aeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeae aeae 1000 ae ae ae ae ae aeaeaeaeaeaeae aeaeaeae aeae aeaeae ae aeae 3.5 4.0 4.5

References

[1] James Deese . The structure of associations in language and thought . Johns Hopkins University Press, 1966 .

[2] Pe ´ter Hala´csy, Andra´s Kornai, La´szlo´ Ne´meth, Andra´s Rung, Istva´n Szakada´t, and Viktor Tro´n. Creating open language resources for hungarian . In LREC , 2004 .

[3] Zsolt Lengyel . Magyar asszocia´cio´s norma´k enciklope´dia´ja I-II. TINTA ko¨nyvkiado´ , Budapest, 2008 - 2010 .

[4] Katherine Nelson . The syntagmaticparadigmatic shift revisited: a review of research and theory . Psychological Bulletin , 84 ( 1 ): 93 , 1977 .

[5] Charles Egerton Osgood . Cross-cultural universals of affective meaning . University of Illinois Press, 1975 .