A statistical exploration of the effects of phoneme class contacts in German noun-noun compounds

Annelen B unne
Leibniz Ins i u e o Ge man Language, Mannheim, GERMANY
Wo king pape : A s a is ical explo a ion o he e ec s o phoneme class
con ac s in Ge man noun-noun compounds
Ve sion: 2025/05/23
This wo king pape desc ibes a s udy ha explo es phoneme con ac p e e ences in Ge man
noun-noun compounds wi h s a is ical me hods on he basis o a la ge da ase (o e 707,000
compounds). The ollowing hypo heses a e es ed based on su p isal alues: 1) owel con ac s
a e a oided  con i med; 2) con ac s o phonemes om he same phoneme class a e a oided
 no con i med; 3) con ac s whe e a phoneme wi h highe sono i y is ollowed by a
phoneme wi h lowe sono i y a e a oided (syllable con ac law)  la gely con i med.
Acknowledgmen s: This wo king pape was de eloped in discussion wi h Alexande
Koplenig, Ka in Hein and S e an Engelbe g, all om Leibniz Ins i u e o Ge man Language,
Mannheim, GERMANY
Supplemen al ma e ial o his wo king pape is a ailable ia OSF: h ps://os .io/e7a d
Con ac add ess: [email p o ec ed]
1 Resea ch ques ions
In his s udy, we examine he phoneme class con ac s ha occu in a la ge numbe o Ge man
noun-noun compounds comp ising wo simplex nouns and look o indica ions whe he
ce ain phonological con ac s a e p e e ed o a oided in compounding. We will ocus on he
ollowing aspec s:
1. Sono i y: Fo syllable con ac s wi hin he same mo phological wo d, phoneme con ac
whe e a mo e sono ous phoneme is ollowed by a less sono ous phoneme a e
p e e ed, while he opposi e is a oided (‘syllable con ac law’; c . e.g. Hall 2011, p.
230-236; Vennemann 1988, p. 40). Do we ind e idence ha his ule is also ele an
o con ac s be ween he cons i uen s in compounding?
2. Vowel con ac s (hia us): Di ec con ac be ween owels is some imes a oided, e.g. in
he con ex o de i a ion (e.g. ame ika-n-isch, ge-g-essen). Can we obse e such a
end in compounding?
3. Con ac o phonemes om he same phoneme class: Con ac s o iden ical phonemes
end o be a oided. As we don’ ha e coding o indi idual phonemes in ou da a, we
will add ess he mo e gene al case o con ac be ween phoneme classes and check
whe he we can obse e any endencies ha his is a oided.
2 Da a se
Ou da a is ex ac ed om he “KoG a Un e suchungsko pus”
1
, which comp ises oughly 7
billion okens and is a subse o he Ge man Re e ence Co pus (DeReKo, Release 2017-II
2
).
This co pus consis s mainly o Ge man newspape ex s (o e 90%), bu also con ains some
li e a y ex s, and abou 6% spoken language ma e ial (c . Bubenho e /Konopka/Schneide
2014). De ailed mo phological in o ma ion was added wi h a cus om wo d analyze based on
he Canoo Language Tools
3
. This made i possible o au oma ically ex ac a la ge collec ion
o o e 489 million nominal compound okens ha se es as he basis o ou s udies on wo d
o ma ion.
Fo his s udy, we used he same da ase as in B unne /Engelbe g/Hein 2021 (s udies A, B and
C). I consis s only o compounds comp ising o wo simplex nouns, e.g. S ad halle (‘ own
hall’). In addi ion o ha , he cons i uen s o he da a se we e anno a ed wi h ca ego ies om
Ge maNe (Hamp and Feldweg 1997; Hen ich and Hin ichs 2010) and we kep only
compounds o which a Ge maNe ma ch was ound o each cons i uen . This had a cleanup
e ec , as i made is mo e likely ha he compounds consis ed o wo plausible wo ds. As
he e we e s ill e o s due o he au oma ic anno a ion, some addi ional manual cleanup was
pe o med on he esul ing lis , emo ing compounds ha con ained cons i uen s ha we e
ei he no ecognizable as wo ds o we e no simplex nouns. The esul ing da ase comp ises
707,910 compound ypes.
4
The au oma ic ool ga e us a segmen a ion o each compound, bu only lemma ized o ms o
he cons i uen s. Fo a phonological s udy howe e , we need su ace o ms, so we in e ed he
segmen a ion o he compound su ace au oma ically, using he a ailable in o ma ion abou
he lemma ized o ms. This p ocess was success ul o 704,475 compounds, 99.5 pe cen o
he o iginal da a se .
5
This is he da a se his s udy is based on. A manual check o 200
andomly selec ed compounds e ealed ha 95% o he assigned su ace segmen a ions we e
1
Ko pus des P ojek s Ko pusg amma ik. Leibniz-Ins i u ü Deu sche Sp ache: „Ko pusges ü z e G amma ik“.
G amma isches In o ma ionssys em g ammis. DOI: 10.14618/ko pusg amma ik. URL: h ps://g ammis.ids-
mannheim.de/ko pusg amma ik/6615.
2
Deu sches Re e enzko pus / A chi de Ko po a gesch iebene Gegenwa ssp ache 2017-II (Release:
01.10.2017). Mannheim: Leibniz-Ins i u ü Deu sche Sp ache. www.ids-mannheim.de/DeReKo.
3
h p://www.canoone .eu. Un o una ely, his websi e no longe exis s. Pa s o he con en o canoone we e
in eg a ed in o LEOdic (h ps://dic .leo.o g/pages/abou /ende/canoone _de.h ml; accessed 23.05.2025) bu he
mo phological analyze is no longe accessible.
4
Fo a mo e de ailed explana ion on how his da ase was cu a ed, c . B unne /Engelbe g/Hein 2021 p. 9-12.
5
Cases whe e su ace o ms could no be assigned we e due o e o s in he au oma ic lemma iza ion o he
cons i uen s, especially he head wo d.
co ec . In he sample, all e o s we e due o p oblems in he o iginal da a, no he pos -
p ocessing s ep: Ei he he inpu da a con ained aul y segmen a ions o he en ies we e no
alid compounds a all.
Any linking elemen s we e ea ed as pa o he i s cons i uen . So o example,
S aa sanwal will be segmen ed in o s aa s (S aa plus linking elemen s) and anwal . We
ha e he in o ma ion o which compounds a simple combina ion o he lemma ized o ms o
he cons i uen s is iden ical o he su ace o m (e.g. A ikabild – A ika plus Bild). This sub-
g oup con ains no linking elemen s and was es ed sepa a ely in ou analyses (c . sec ion 3.2,
end).
To s udy phoneme con ac in compounding we needed in o ma ion abou he las phoneme o
he modi ie (including linking elemen ) and he i s phoneme o he head o each
compound. We de ised a ule-based sys em o de i e he phoneme classes based on he
o hog aphical su aces, iden i ying eigh phoneme classes ha a e anked by sono i y. Table
1 shows he labels used o he phoneme classes. They a e comp ised o a le e as sho hand
o he name o he class and a numbe ha indica es he le el o sono i y wi h 1 being he
leas sono ous and 8 he mos sono ous phoneme class. Appendix 1 documen s he exac ules
ha we e applied o assign labels. No e ha in his s udy, he app oximan /semi owel [j] ( i s
phoneme o e.g. Jah ) is classi ied as a high owel and [ɐ] ( he las phoneme o e.g. Lebe ) is
conside ed a low owel. The glo al s op [ʔ] is no conside ed a phoneme bu a phone ic
ea u e because in S anda d Ge man i appea s p edic ably in ce ain con ex s (wo d-ini ial o
s essed syllables ha begin wi h a owel).
Table 1: Labels o he phoneme classes, anked by sono i y (examples a e no exhaus i e)
label
phoneme class
examples
1_p
plosi e
[ ] [g] [k] [b] [d] [p]
2_
ica i e
[s] [ ] [ʃ] [h] [x]
3_n
nasal
[n] [m] [ŋ]
4_l
liquid (phoneme l)
[l] [l]
5_l
liquid (phoneme )
[ʁ] [ʀ] [ ]
6_
high owel
[ɪ] [ʊ] [i] [ʏ] [i] [u] [y] [j]
7_
mid owel
[ə] [ɛ] [ɔ] [œ] [e] [o]
8_
low owel
[a] [aː] [ɐ]
I mus be no ed ha ou da a se con ains a ela i ely high numbe o p ope names as well as
o eign language elemen s, especially English and F ench cons i uen s. Though we ied o
accoun o o eign languages in ou ules o some ex en , his led o some ambigui ies. A
manual check o a sample o 200 compounds e ealed ha he assignmen o he phoneme
class was co ec in 97% pe cen o he cases. The e o s we e all due o o eign language
cons i uen s. Table 2 shows an exce p om he da ase .
Table 2: Exce p om he da a se ( o al size: 704,475 en ies)
equency
compound
modi ie
su ace
head
su ace
phoneme class
end o modi ie
phoneme class
beginning o head
2911504
Sonn ag
sonn
ag
3_n
1_p
1758319
Donne s ag
donne s
ag
2_
1_p
1269620
Bü ge meis e
bü ge
meis e
5_l
3_n
1160962
Wochenende
wochen
ende
3_n
7_
781104
Gebu s ag
gebu s
ag
2_
1_p
696116
Fußball
uß
ball
2_
1_p
467966
Land ag
land
ag
1_p
1_p
465068
Landk eis
land
k eis
1_p
1_p
449355
Kinde ga en
kinde
ga en
5_l
1_p
428906
Jah zehn
jah
zehn
5_l
1_p
3 Empi ical explo a ion
The e a e wo possibili ies o use he quan i a i e da a: Ei he we only look a he ypes in ou
da a se o we also inco po a e he equency in o ma ion we ha e o hese ypes. Bo h
app oaches ha e me i : Conside ing only he ypes gi es us an idea which dis inc compounds
occu in eal li e da a. Howe e , as each compound ype has he same impac on he esul ,
e y in equen compounds and e o s in he da a gain s ong in luence. I we weigh he ypes
acco ding o hei equencies, we ge in o ma ion abou which dis inc compounds a e
common, which is a guably mo e use ul o ell us some hing abou p oduc i i y and
en enchmen in Ge man compounding. Howe e , as he equency cu e in ou da a se
ollows a Zip -like dis ibu ion wi h ew e y equen en ies and a huge ail end o hapax
legomena, he mos equen en ies gain a s ong in luence on he esul s (c. . igu e 1). As
bo h aspec s a e in e es ing, we will look a he dis ibu ions on bo h ways and discuss he
esul s.
Figu e 1: F equency cu e o he compounds in ou da a se (loga i hmic scale)
3.1 Da a dis ibu ion ma ices
In he ollowing ma ices ( igu es 2-5), he phoneme classes o he las phoneme o he
modi ie s a e plo ed on he y-axis and he phoneme classes o he i s phoneme o he head
wo ds on he x-axis. Each scale is o de ed acco ding o sono i y (leas o mos sono ous). The
cells show he alues o con ac s be ween hose phoneme classes. Fo example, igu e 1
shows ha a con ac be ween a ica i e as las phoneme o he modi ie and a plosi e as i s
phoneme o he head (sho hand: [2_ + 1_p]) occu s in 47,751 compounds ypes. The cells
o ming he diagonal line om uppe le o lowe igh ep esen he cases whe e phonemes
o he same class come in o con ac . I he hypo hesis holds ue ha con ac s whe e he
second elemen is less sono ous han he i s a e p e e ed, he cells below his diagonal line
should ha e gene ally highe alues han he cells abo e i . The 9 cells in he lowe igh
quad an o he ma ix show he con ac s be ween owel phoneme classes. I he hypo hesis
ha owel con ac s a e a oided is ue, hose should ha e lowe alues.
Figu e 2 shows a ma ix o ype coun s o all combina ions o phoneme class con ac s. I is
e iden ha some phoneme classes, e.g. plosi es, a e much mo e common han o he s. To ge
an imp ession whe he he e a e dis ibu ional p e e ences besides simple equency o he
phoneme classes, we calcula ed he numbe o cases ha would appea in each ma ix cell, i
he dis ibu ion we e andom – he expec ed dis ibu ion.
6
Fo igu e 3, we calcula ed which
pe cen age o he expec ed alues he eal alues ep esen . I a eal numbe co esponded
exac ly o he expec ed numbe , he cell alue would be 100. Looking a he con ac [2_ +
1_p] again, we see ha he alue is 102.83. This means ha i is 102.83% o he expec ed
alue, i.e. only sligh ly la ge han expec ed.
Figu e 3 indica es ha he de ia ions om he expec ed numbe s a e no la ge, he cell alues
ange be ween 73 and 111. The dis ibu ion o con ac s be ween phoneme classes in
compounds seems qui e uni o m on ype le el. The e migh be a sligh end ha con ac s
wi h dec easing sono i y a e a o ed, which would suppo ou hypo hesis, bu he pic u e is
no clea .
We can, howe e , obse e a end ha con ac s be ween owels a e a oided, especially hose
whe e owels o same phoneme class come in o con ac and hose whe e he second owel is
mo e sono ous.
When looking a con ac s be ween iden ical phoneme classes o any kind, we see ha hese
a e a li le less equen han expec ed as well.
6
The e a e s a is ical es s o de e mine whe he he dis ibu ion in a ma ix de ia es om he expec ed
dis ibu ion in a signi ican way; a common one is he Chi squa e es . I we un i o he ma ices in igu es 2
and 4, we do ge signi ican esul s – so we know ha he dis ibu ion is no andom. Howe e , when applied
o a e y la ge da ase such as ou s, he s a is ical es can be misleading, as i picks up on small e ec s. In
addi ion o ha , he assump ion o independence unde lying he s a is ical es s does no hold o language
da a in gene al, which makes i e en mo e p oblema ic o ely on he es esul . We he e o e op ed o a
pu ely desc ip i e app oach he e.

Figu e 2: Dis ibu ion o compound ypes acco ding o hei phoneme con ac s ( ype-based)
Figu e 3: Pe cen age o eal equencies in ela ion o expec ed equencies ( ype-based)
We will now look a he oken-based dis ibu ion, aking in o accoun he equency o each
compound ype ( igu e 4 and 5). The de ia ions om he expec ed dis ibu ion a e much
la ge – while in igu e 2 he alues anged only be ween 73 and 111, in igu e 5 hey ange
om ca. 38% (less han hal he expec ed alue) o nea ly 300 ( h ee imes mo e han
expec ed). We again obse e a sligh p e e ence o con ac s whe e he second phoneme is
less sono ous and a disp e e ence o mos owel combina ions, bu he e a e ou lie s.
As men ioned abo e, he equency cu e o ou da a is Zip -like wi h a e y s eep slope (c .
igu e 1). I we ake a close look a a ew cells ha s and ou in pa icula wi h much highe
alues han expec ed, we see ha compounds om he highes equency bands ha e indeed
an impac : The combina ion [3_n + 7_ ] (nasal + mid owel) has he highes alue wi h
294.55%. Compounds wi h his combina ion include Wochenende which is one o he mos
equen compounds in ou da a. Simila ly, he combina ion [3_l + 6_ ] (liquid + high owel),
which s ands ou wi h a alue o 274.44 %, includes he e y equen wo d Schuljah .
Jus emo ing he wo ds Wochenende and Schuljah al eady leads o a ma ix whe e he
espec i e cell alues a e much less ex eme, bu in o de o ha e a mo e objec i e pic u e o
he in luence o e y equen compounds, we ecalcula ed he ma ix wi hou he 100 mos
equen en ies ( igu e 6). The dis ibu ion becomes no ably mo e e en. The ma ix now
looks mo e simila o he ype-based ma ix ( igu e 3), albei wi h s onge de ia ions om he
expec ed dis ibu ion in bo h di ec ions. Only he cell [6_ + 4_l] (high owel + liquid) s ill
has he alue 234,72, i.e. mo e han wice he expec ed i ems. This cell con ains equen
compounds as well (e.g. EU-Land, Indus ieland), bu is no domina ed by a single en y.
Figu e 4: Dis ibu ion o compound okens acco ding o hei phoneme con ac s
Figu e 5: Pe cen age o he eal numbe s in ela ion o he expec ed numbe s ( okens)
Figu e 6: Pe cen age o he eal numbe s in ela ion o he expec ed numbe s ( okens wi hou
he 100 mos equen compounds)
As men ioned a he beginning o he sec ion, he plo s shown he e a e based on he oken
coun s in ou da a. We ecalcula ed he same plo s omi ing he 100 mos equen compounds
o con ol o he e ec o e y equen compounds (c . igu e 6). Though he boxplo s look
sligh ly di e en , he pe mu a ion es s ga e he same esul s. We also calcula ed he plo s
based on ype coun s (c . igu es 2 and 3). Again, he pe mu a ion es s we e signi ican o
he same g oup con as s.
In sec ion 2 we epo ed ha linking elemen s we e coun ed as pa o he modi ie . This
implies ha in cases whe e a linking elemen is p esen , he las phoneme o his linking
elemen was used in ou s udy. As linking elemen s a e a special ( hough common) case in
compounding, we epea ed ou calcula ions excluding all compounds whe e linking elemen s
we e de ec ed. S ill, we could con i m he same signi ican di e ences on he basis o okens,
okens wi hou he 100 mos equen compounds and on basis o ypes.
To sum up, he ollowing esul s p o ed s able o di e en con igu a ions:
 Con ac s be ween owel classes ha e signi ican ly highe su p isal alues.
 Con ac s ha a e ‘bad’ acco ding o he syllable con ac law ha e signi ican ly highe
su p isal alues han hose ha a e ega ded as ‘good’.
 The e we e no signi ican esul s o he o he es s.

4 Appendix
4.1 Rules o assigning phoneme classes
The ollowing ules we e applied o assign a phoneme class o he o hog aphical o m o he
las elemen o he modi ie and he i s elemen o he head wo d. The s a egy was he same
in bo h asks: We looked a he end o he wo d ( o modi ie s) and a he beginning o he
wo d ( o heads) and checked longe le e combina ions i s . Each able lis s he ules in he
o de in which hey we e applied o he o hog aphical su ace. I a ule ma ched, execu ion
was s opped and he co esponding phoneme class label was assigned.
No e ha some ules in hese ables a e qui e idiosync a ic o ou da ase as we had o
accoun o some o eign, especially English and F ench spellings and decided o include
explici ules o special cases o deal wi h ambigui y.
Rules o he phoneme class o he las phoneme o he compound modi ie
execu ion o de
le e combina ions
label
class name
1
["aille", " iew"]
6_
high owel
2
["sch", "löw"]
2_
ica i e
3
["ieh"]
6_
high owel
4
[" h", "qu", "ig"]
1_p
plosi e
5
["ch", "sh", "ph", "p "]
2_
ica i e
6
["ng"]
3_n
nasal
7
[" h"]
5_l
liquid ( )
8
["ie", "ih", "uh", "ou", "üh", "ei", "ai", "eu",
"äu", "au", "ew"]
6_
high owel
9
["eh", "oh", "äh", "öh", "ow", "aw"]
7_
mid owel
10
["ah"]
8_
low owel
11
[" ", "d", "g", "k", "b", "c", "p"]
1_p
plosi e
12
["ß", "z", "s", " ", " ", "h", "x"]
2_
ica i e
13
["n", "m"]
3_n
nasal
14
["l"]
4_l
liquid (l)
15
[" "]
5_l
liquid ( )
16
["i", "u", "y", "ü", "j", "q10"]
6_
high owel
17
["e", "o", "ä", "ö", "é"]
7_
mid owel
18
["a", "à"]
8_
low owel
Rules o he phoneme class o he i s phoneme o he compound head
execu ion o de
le e combina ions
label
class name
1
["sch"]
2_
ica i e
2
["ch", "sh", "ph"]
2_
ica i e
3
["ei"]
8_
high owel
4
[" ", "d", "g", "k", "b", "x", "p", "c", "q", "z"]
1_p
plosi e
5
["s", " ", " ", "w", "h"]
2_
ica i e
6
["m", "n"]
3_n
nasal
7
["l"]
4_l
liquid (l)
10
“q” igge s an anno a ion as 6_ (high owel) because in ou da a se he only modi ie ha ended in he
le e “q” was “IQ”. As an abb e ia ion his is p onounced wi h an “u” sound a he end.
8
[" "]
5_l
liquid ( )
9
["i", "u", "ü", "j", "y"]
6_
high owel
10
["e", "o", "ä", "ö"]
7_
mid owel
11
["a"]
8_
low owel
4.2 Table o su p isal alues, based on compound okens
Addi ional ables o o he da a con igu a ions a e a ailable a he OSF eposi o y.
Modi ie
(las
phoneme)
Head
( i s
phoneme)
F equency
o
combina ion
To al
equency
o
modi ie
class
p(head
class|
modi ie
class)
su p isal
same
class
con ac
owel
class
con ac
sono i y
sono i y
di e ence
1_p
1_p
9782346
26229355
0,372954
1,42293
yes
no
same
0
1_p
2_
10066856
26229355
0,383801
1,381569
no
no
mo e
-1
1_p
3_n
2236945
26229355
0,085284
3,551581
no
no
mo e
-2
1_p
4_l
529642
26229355
0,020193
5,630021
no
no
mo e
-3
1_p
5_l
1230246
26229355
0,046903
4,414164
no
no
mo e
-4
1_p
6_
334875
26229355
0,012767
6,291416
no
no
mo e
-5
1_p
7_
696491
26229355
0,026554
5,234934
no
no
mo e
-6
1_p
8_
1351954
26229355
0,051544
4,278064
no
no
mo e
-7
2_
1_p
9713135
17242357
0,56333
0,827948
no
no
less
1
2_
2_
3886129
17242357
0,225383
2,149551
yes
no
same
0
2_
3_n
1657100
17242357
0,096106
3,379224
no
no
mo e
-1
2_
4_l
430253
17242357
0,024953
5,324628
no
no
mo e
-2
2_
5_l
195860
17242357
0,011359
6,45999
no
no
mo e
-3
2_
6_
122602
17242357
0,007111
7,135831
no
no
mo e
-4
2_
7_
212057
17242357
0,012299
6,345361
no
no
mo e
-5
2_
8_
1025221
17242357
0,059459
4,07195
no
no
mo e
-6
3_n
1_p
8074539
16724684
0,482792
1,050527
no
no
less
2
3_n
2_
4626159
16724684
0,276607
1,854092
no
no
less
1
3_n
3_n
674675
16724684
0,04034
4,631642
yes
no
same
0
3_n
4_l
463551
16724684
0,027717
5,173107
no
no
mo e
-1
3_n
5_l
373353
16724684
0,022323
5,485295
no
no
mo e
-2
3_n
6_
176727
16724684
0,010567
6,564313
no
no
mo e
-3
3_n
7_
1444712
16724684
0,086382
3,533125
no
no
mo e
-4
3_n
8_
890968
16724684
0,053273
4,230462
no
no
mo e
-5
4_l
1_p
3251965
7325810
0,443905
1,171677
no
no
less
3
4_l
2_
2355677
7325810
0,321559
1,636847
no
no
less
2
4_l
3_n
516733
7325810
0,070536
3,825497
no
no
less
1
4_l
4_l
181935
7325810
0,024835
5,331493
yes
no
same
0
4_l
5_l
211360
7325810
0,028851
5,115214
no
no
mo e
-1
4_l
6_
290147
7325810
0,039606
4,658132
no
no
mo e
-2
4_l
7_
164437
7325810
0,022446
5,477381
no
no
mo e
-3
4_l
8_
353556
7325810
0,048262
4,372978
no
no
mo e
-4
5_l
1_p
5547747
13783101
0,402504
1,312927
no
no
less
4
5_l
2_
4131515
13783101
0,299752
1,738158
no
no
less
3
5_l
3_n
2476063
13783101
0,179645
2,476781
no
no
less
2
5_l
4_l
486991
13783101
0,035332
4,822862
no
no
less
1
5_l
5_l
300124
13783101
0,021775
5,521198
yes
no
same
0
5_l
6_
168566
13783101
0,01223
6,353443
no
no
mo e
-1
5_l
7_
203811
13783101
0,014787
6,079525
no
no
mo e
-2
5_l
8_
468284
13783101
0,033975
4,879373
no
no
mo e
-3
6_
1_p
1159659
2641025
0,439094
1,187397
no
no
less
5
6_
2_
880392
2641025
0,333352
1,58488
no
no
less
4
6_
3_n
188100
2641025
0,071222
3,811526
no
no
less
3
6_
4_l
169247
2641025
0,064084
3,963896
no
no
less
2
6_
5_l
69348
2641025
0,026258
5,2511
no
no
less
1
6_
6_
30441
2641025
0,011526
6,438938
yes
yes
same
0
6_
7_
31913
2641025
0,012084
6,37081
no
yes
mo e
-1
6_
8_
111925
2641025
0,042379
4,560494
no
yes
mo e
-2
7_
1_p
1724384
4290094
0,401946
1,314928
no
no
less
6
7_
2_
1399232
4290094
0,326154
1,616374
no
no
less
5
7_
3_n
640368
4290094
0,149267
2,744036
no
no
less
4
7_
4_l
148193
4290094
0,034543
4,85546
no
no
less
3
7_
5_l
119316
4290094
0,027812
5,16815
no
no
less
2
7_
6_
61574
4290094
0,014353
6,122544
no
yes
less
1
7_
7_
43610
4290094
0,010165
6,620206
yes
yes
same
0
7_
8_
153417
4290094
0,035761
4,805479
no
yes
mo e
-1
8_
1_p
491683
1028146
0,478223
1,064245
no
no
less
7
8_
2_
273897
1028146
0,266399
1,90834
no
no
less
6
8_
3_n
137226
1028146
0,133469
2,905419
no
no
less
5
8_
4_l
32282
1028146
0,031398
4,993171
no
no
less
4
8_
5_l
39380
1028146
0,038302
4,706438
no
no
less
3
8_
6_
15351
1028146
0,014931
6,065569
no
yes
less
2
8_
7_
12613
1028146
0,012268
6,34899
no
yes
less
1
8_
8_
25714
1028146
0,02501
5,321347
yes
yes
same
0
Bibliog aphy
B unne , A., Engelbe g, S. & K. Hein. 2021. The dis ibu ion o cons i uen wo ds in nominal
compounds and i s impac on seman ic in e p e a ion: an empi ical s udy. In: Jou nal o Wo d
Fo ma ion 1: 7-36.
Bubenho e , N., Konopka, M. & R. Schneide (Eds.). 2014. P älimina ien eine
Ko pusg amma ik. Tübingen: Na .
Gibson, E., Fu ell, R., Pian adosi S. P., Dau iche, I., Mahowald, K., Be gen, L. & R. Le y.
2019. How E iciency Shapes Human Language. T ends in cogni i e sciences 23 (5): 389-
407.
Hall, T. Alan. 2011. Phonologie. Eine Ein üh ung. Be lin / New Yo k: De G uy e .
Hamp, B. & Feldweg, H. 1997. Ge maNe – a Lexical-Seman ic Ne o Ge man. In
P oceedings o he ACL wo kshop Au oma ic In o ma ion Ex ac ion and Building o Lexical
Seman ic Resou ces o NLP Applica ions. Mad id 1997. 9–15.
Hen ich, V. & Hin ichs, E. 2010. Ge nEdiT – The Ge maNe Edi ing Tool. In P oceedings o
he Se en h Con e ence on In e na ional Language Resou ces and E alua ion (LREC 2010),
2228–2235. Valle a, Mal a.
Koplenig, A. 2019. A non-pa ame ic signi icance es o compa e co po a. PLoS ONE 14(9).
[online: h ps://doi.o g/10.1371/jou nal.pone.0222703]
Koplenig, A., Wol e , S., Rüdige , J. O., & Meye , P. 2024. Human languages ade o
complexi y agains e iciency. In: Cha lo es ille: OSF P ep in s. [online:
h ps://os .io/p ep in s/os /8xgqz_ 1]
Shannon, C. E. 1948. A Ma hema ical Theo y o Communica ion. Bell Sys em Technical
Jou nal 27: 379-423.
Tu, N. D. T.. 2024. Eine ko puslinguis ische Un e suchung zu lexikalischen Viel al on
di ek en und indi ek en Redeeinlei e n. Mannheim: IDS-Ve lag. [online: h ps://pub.ids-
mannheim.de/lau end/idsopen/idsopen06.h ml]
Vennemann, T. 1988. P e e ence Laws o Syllable S uc u e and he Explana ion o Sound
Change. Be lin / New Yo k: De G uy e .