White Paper Series

THE CROATIAN
LANGUAGE
IN THE
DIGITAL AGE

Niz Bijele Knjige

HRVATSKI
JEZIK U
DIGITALNOM
DOBU
Marko Tadi
Dunja Brozovi-Ronevi
Amir Kapetanovi

White Paper Series

THE CROATIAN
LANGUAGE
IN THE
DIGITAL AGE

Niz Bijele Knjige

HRVATSKI
JEZIK U
DIGITALNOM
DOBU
Marko Tadi [1]
Dunja Brozovi-Ronevi
Amir Kapetanovi [2]

[2]

[1] Filozofski Fakultet, Zagreb

[2] Institut za hrvatski jezik i jezikoslovlje

Georg Rehm, Hans Uszkoreit

(urednici, editors)

PREDGOVOR PREFACE
Ova bijela knjiga dio je niza koji promie jezine teh-

is white paper is part of a series that promotes

nologije i njihove mogunosti. Namijenjena je novi-

knowledge about language technology and its poten-

narima, politiarima, jezinim zajednicama, uiteljima,

tial. It addresses journalists, politicians, language com-

predavaima i ostalima. Dostupnost i uporaba jezinih

munities, educators and others. e availability and

tehnologija u Europi razliita je od jezika do jezika. Su-

use of language technology in Europe varies between

sljedno, razliite su i aktivnosti potrebne za daljnju pot-

languages. Consequently, the actions that are required

poru istraivanjima i razvoju jezinih tehnologija od je-

to further support research and development of lan-

zika do jezika. Potrebne akcije ovise o mnogo imbe-

guage technologies also diers. e required actions

nika kao to su sloenost pojedinoga jezika i veliina

depend on many factors, such as the complexity of a

dotine jezine zajednice.

given language and the size of its community.

Mrea izvrsnosti META-NET, koju podupire Europ-

META-NET, a Network of Excellence funded by the

ska komisija, provela je analizu trenutano raspoloi-

European Commission, has conducted an analysis of

vih jezinih resursa i tehnologija u ovome nizu bije-

current language resources and technologies in this

lih knjiga (s. 93). Ta je analiza usredotoena ponaj-

white paper series (p. 93). e analysis focused on the

prije na 23 slubena jezike Europske unije, ali i na os-

23 ocial European languages as well as other impor-

tale vane nacionalne i regionalne jezike u Europi. Re-

tant national and regional languages in Europe. e re-

zultati ove analise ukazuju na nesrazmjerne nedostatke

sults of this analysis suggest that there are tremendous

u tehnolokoj potpori i znaajne istraivake nedos-

decits in technology support and signicant research

tatke za svaki od promatranih jezika. Predstavljena po-

gaps for each language. e given detailed expert anal-

drobna struna analiza i procjena trenutane situacije

ysis and assessment of the current situation will help

pomoi e u uinkovitosti dodatnih istraivanja u tome

maximise the impact of additional research.

smjeru.

As of November 2011, META-NET consists of 54

Od mjeseca studenoga 2011. META-NET se sastoji od

research centres from 33 European countries (p. 89).

54 istraivaka sredita iz 33 europske zemlje (s. 89).

META-NET is working with stakeholders from econ-

META-NET surauje s kljunim dionicima iz gospo-

omy (soware companies, technology providers, users),

darstva (tvrtke koje izgrauju programsku podrku,

government agencies, research organisations, non-

tehnoloki isporuitelji, korisnici), vladinim agenci-

governmental organisations, language communities

jama, istraivakim organizacijama, nevladinim orga-

and European universities. Together with these com-

nizacijama, jezinih zajednicama i europskim sveuili-

munities, META-NET is creating a common technol-

tima. Zajedno s njima META-NET stvara zajedniku

ogy vision and strategic research agenda for multilin-

tehnoloku viziju i strateki plan za viejezinu Europu

gual Europe 2020.

2020.

III

META-NET oce@meta-net.eu http://www.meta-net.eu

Autori ovoga dokumenta zahvalni su autorima Bijele knjige o

njemakome jeziku za doputenje uporabe odabrane jezinoneovisne grae iz njihovoga teksta [1].

e authors of this document are grateful to the authors of

the White Paper on German for permission to re-use selected
language-independent materials from their document [1].

Izradba ove bijele knjige poduprta je od strane Sedmoga okvirnoga programa i ICT programa za podrku politici Europske
komisije u skladu s ugovorima T4ME (opi ugovor 249 119),
CESAR (opi ugovor 271 022), METANET4U (opi ugovor
270 893) i META-NORD (opi ugovor 270 899).

e development of this white paper has been funded by the

Seventh Framework Programme and the ICT Policy Support
Programme of the European Commission under the contracts
T4ME (Grant Agreement 249 119), CESAR (Grant Agreement 271 022), METANET4U (Grant Agreement 270 893)
and META-NORD (Grant Agreement 270 899).

IV

SADRAJ CONTENTS
HRVATSKI JEZIK U DIGITALNOM DOBU
1 Saetak

2 Jezici u opasnosti: izazov za jezine tehnologije

2.1

Jezine granice koe europsko informacijsko drutvo . . . . . . . . . . . . . . . . . . . . . . . .

2.2

Opasnost za nae jezike . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.3

Jezine su tehnologije kljune potporne tehnologije . . . . . . . . . . . . . . . . . . . . . . . . .

2.4

Mogunosti jezinih tehnologija . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.5

Izazovi koji stoje pred jezinim tehnologijama . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.6

Usvajanje jezika kod ljudi i strojeva . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 Hrvatski jezik u europskome informacijskome drutvu

3.1

Ope injenice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.2

Hrvatska narjeja . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3.3

Standardizacija hrvatskoga jezika . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3.4

Osobine hrvatskoga jezika . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3.5

Odnos hrvatskoga standardnoga jezika s ostalim jezicima tokavske osnovice . . . . . . . . . . . 18

3.6

Skrb o jeziku u Hrvatskoj . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

3.7

Jezik u obrazovanju . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3.8

Meunarodni odnosi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.9

Hrvatski na Internetu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

4 Jezinotehnoloka podrka za hrvatski

23

4.1 Arhitekture jezinotehnolokih aplikacija . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

4.2 Osnovna podruja primjene jezinih tehnologija . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3 Jezine tehnologije u obrazovanju . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.4 Nacionalni projekti i inicijative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.5 Dostupnost alata i resursa za hrvatski jezik . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.6 Usporedba izmeu jezika . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.7 Zakljuci . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

5 O META-NET-u

42

THE CROATIAN LANGUAGE IN THE DIGITAL AGE

1 Executive Summary

43

2 Languages at Risk: a Challenge for Language Technology

45

2.1

Language Borders Hold back the European Information Society . . . . . . . . . . . . . . . . . . 46

2.2

Our Languages at Risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

2.3

Language Technology is a Key Enabling Technology . . . . . . . . . . . . . . . . . . . . . . . . 47

2.4

Opportunities for Language Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

2.5

Challenges Facing Language Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

2.6

Language Acquisition in Humans and Machines . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

3 The Croatian Language in the European Information Society

50

3.1

General Facts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

3.2

Croatian dialects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

3.3

Standardisation of Croatian language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

3.4

Characteristics of the Croatian language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

3.5

The Croatian standard language and other tokavian-structured languages . . . . . . . . . . . . 60

3.6

Linguistic cultivation in Croatia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

3.7

Language in education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

3.8

International aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

3.9

Croatian on the Internet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

4 Language Technology Support for Croatian

65

4.1 Application Architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

4.2 Core application areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3 Educational programmes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.4 National projects and initiatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.5 Availability of tools and resources for Croatian . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.6 Cross-language comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

5 About META-NET

83

A Bibliograja -- References

85

B META-NET lanice -- META-NET Members

89

C Niz Bijele Knjige META-NET -- The META-NET White Paper Series

93

1
SAETAK
Informacijske tehnologije mijenjaju na svakodnevni i-

cima u globaliziranome informacijskome drutvu, nego

vot. Svakodnevno se sluimo raunalima za pisanje, ure-

i promjenu njegovih sociolingvistikih okolnosti koja se

ivanje, raunanje, pretragu obavijesti i sve vie za i-

moe oekivati u 2013. kad e postati 24. slubeni je-

tanje, sluanje glazbe, pregledavanje fotograja i gleda-

zik Europske unije. Od toga trenutka za hrvatski se jezik

nje lmova. U svojim depovima nosimo mala raunala

oekuje dostupnost itavoga niza jezinotehnolokih re-

koja koristimo za obavljanje telefonskih poziva, pisanje

sursa, alata i usluga kakve ve postoje, ali se isto tako i

e-pote, prikupljanje obavijesti i za zabavu gdje god se

dalje nesmetano razvijaju za ostale slubene jezike EU-a.

nalazili. Kako ta masovna digitalizacija obavijesti, zna-

Trailice koje mogu pretraivati puni tekst prema svim

nja i svakodnevnih komunikacija utjee na na jezik?

oblicima u kojima se hrvatske rijei mogu pojavljivati,

Hoe li se na jezik promijeniti ili ak nestati? Kakve

sustavi za diktiranje tj. automatsko pretvaranje govora

su mogunosti hrvatskoga jezika za preivljavanje?

na hrvatskome u tekst, ili, moda najvaniji, sustavi za

Mnogi od est tisua jezika na svijetu ne e preivjeti

strogo prevoenje na i sa hrvatskoga, samo su neki od

u globaliziranom digitalnom informacijskom drutvu.

primjera uporabivosti jezinih tehnologija koje se oe-

Procjenjuje se kako je barem dvije tisue jezika osueno

kuju ne samo kao istraivaki prototipovi, nego i kao

na izumiranje u sljedeem desetljeu. Preostali e nasta-

korisni komercijalni proizvodi. Ne moemo oekivati

viti igrati ulogu u privatnome krugu obitelji ili susjed-

kako e ih za hrvatski jezik izraditi istraivai koji se bave

stva, ali ne nuno i na razini opega poslovanja ili na

engleskim, francuskim, njemakim, ekim, slovenskim

akademskoj razini. Status jezika ne ovisi samo o broju

ili srpskim, ve te jezine resurse, alate i usluge moramo

njegovih govornika ili broju knjiga, lmova i TV-postaja

razviti sami. Meutim, utoliko e nam biti lake ako te

koje se njime slue, nego i o prisutnosti toga jezika u di-

napore uskladimo i koordiniramo sa slinim takvim na-

gitalnome informacijskome prostoru i u adekvatnoj pro-

porima za druge EU jezike, a upravo tome slui inicija-

gramskoj podrci.

tiva opisana u ovoj tiskovini.

U dananjem informacijski usmjerenom drutvu, mogunost dostupa obavijestima na vlastitome jeziku sma-

Ova bijela knjiga o hrvatskome jeziku pokazuje kako u

tra se dosegnutom civilizacijskom razinom nezaobilaz-

Hrvatskoj postoji temeljno okruje za istraivanje jezi-

nom za prevladavaju digitalnoga jaza. Naime, jezine za-

nih tehnologija, meutim to do sada nije rezultiralo i

jednice, koje za svoj jezik ne budu imale razvijene jezine

razvojem jezine industrije. Unato tome to su za hr-

tehnologije, ostat e s druge strane digitalne razdjelnice.

vatski izraeni neki jezini resursi i tehnologije, znatno

Kad je rije o hrvatskome jeziku i jezinim tehnologi-

ih je manje nego za druge slavenske jezike, npr. eki, a

jama, onda ponajprije valja imati na umu ne samo osigu-

jo ih je manje razvijeno u usporedbi s veim europskim

ranje njegova ravnopravnoga sudjelovanja s drugim jezi-

jezicima kao to su engleski, njemaki ili francuski.

Premda u Hrvatskoj postoji ve polustoljetna tradicija

vjeu, potrebno je poduzeti niz ciljanih mjera kako bi se

istraivanja na podruju raunalnoga jezikoslovlja, rau-

hrvatski jezini resursi i alati doveli na istu razinu razvi-

nalne obradbe teksta i korpusne lingvistike (uz nastanak

jenosti glede njihove kakvoe i koliine, kakva je razina

tako znaajnih resursa kao to su Hrvatski estotni rje-

ve dosegnuta za druge europske jezike.

nik, Hrvatski nacionalni korpus, Hrvatsko-engleski us-

Vizija META-NET-a su visokokvalitetne jezine tehno-

poredni korpus, Hrvatski morfoloki leksikon, Hrvat-

logije za sve jezike koje podupiru politiko i gospodarsko

ska ovisnosna banka stabala, itd.), ne moe se rei da

jedinstvo kroz kulturnu raznolikost. Ove e tehnologije

je sadanje stanje jezinih tehnologija zadovoljavajue.

pomoi u uklanjanju prepreka i u izgradnji mostova iz-

Uz nacionalno podupirane projekte, koji su na alost jo

meu jezika u Europi. To, meutim, trai od svih di-

uvijek malobrojni, od 2008. zapoinje se ozbiljnija pot-

onika ovoga procesa politike, istraivanja, gospodar-

pora kroz pet projekata Europske komisije: CLARIN,

stva i drutva u cjelini objedinjavanje svojih napora u

ACCURAT, LetsMT!, ATLAS, XLike; ali i oni su ma-

budunosti.

hom usmjereni na rjeavanje pojedinanih problema ili

Ovaj niz bijelih knjiga nadopunjuje ostale strateke ak-

pruanja tehnolokih rjeenja, a rijetko na ukupnost je-

tivnosti koje poduzima META-NET. Najnovije obavi-

zinih tehnologija za hrvatski jezik. Tu ulogu za hrvat-

jesti, kao to su trenutana inaica vizije META-NET-

ski jezik preuzima esti projekt CESAR kao i ira

a [2] ili Strateki istraivaki plan (SIP) moe se pro-

META-NET inicijativa, stvaranjem ove bijele knjige.

nai na META-NET-ovim mrenim stranicama: http:

Prema procjenama podrobnije iznesenim u ovome iz-

//www.meta-net.eu.

2
JEZICI U OPASNOSTI: IZAZOV ZA JEZINE
TEHNOLOGIJE
U ovome trenutku svjedoimo digitalnoj revoluciji koja

stvaranjem razliitih medija kao to su knjige, no-

korjenito utjee na nau komunikaciju i nae drutvo.

vine, radio, televizija i drugi, zadovoljavaju se komu-

Najnoviji razvoj digitalnih i mrenih komunikacijskih

nikacijske potrebe puanstva.

tehnologija ponekad se usporeuju s Gutenbergovim

izumom tiska pominim slovima. to nam ta analo-

U zadnjih je dvadeset godina informacijska tehnologija

gija moe rei o budunosti europskoga informacijskoga

omoguila olakavanje i automatizaciju mnogih pro-

drutva i o naim vlastitim jezicima?

cesa:
raunalna priprema za tisak zamijenila je tipkanje i

Digitalna revolucija usporediva je s

Gutenbergovim izumom tiska pominim slovima.

graki slog;
Microso PowerPoint zamijenio je projiciranje s
prozirnica;

Nakon Gutenbergova izuma pravi su proboji u komunikaciji i razmjeni znanja postignuti pothvatima kao to je
Lutherov prijevod Biblije na narodni jezik (ili u hrvatskome sluaju, glagoljiki prvotisak Misala iz 1483. kao
prve tiskanje knjige na hrvatskome jeziku). U nadolazeim stoljeima razvijeni su razni kulturni postupci koji
su omoguili obradbu jezika i razmjenu znanja:

e-pota omoguuje odailjanje i primanje dokumenata bre od telefaks ureaja;

Skype nudi jeine internetske telefonske pozive i
odravanje virtualnih sastanaka;
zajedniki formati zapisa zvunih i vizualnih podataka omoguuju jednostavnu razmjenu multimedijskih sadraja;

pravopisno i gramatiko normiranje veih jezika

omoguilo je brzu razmjenu novih znanstvenih
ideja;
uspostavljanje slubenih jezika omoguilo je graanima komu-nikaciju unutar odreenih (esto politikih) granica;
pouavanje jezika i prevoenje omoguilo je raz-

trailice omoguuju pristup www-stranicama na temelju pretrage uporabom kljunih rijei;

mrene usluge poput Google prevoditelja nude brze,
ali zato pribline prijevode;
drutvene mree kao to su Facebook, Twitter i Google+ pospjeuju komunikaciju, omoguuju suradnju i dijeljenje obavijesti.

mjenu preko jezinih granica;

stvaranje urednikih i bibliografskih normi osiguralo je kakvou tiskovina;

Premda su takve aplikacije i usluge viestruko korisne,

ipak jo ne mogu podupirati u cijelosti odrivo, vie-

jezino europsko drutvo u kojem informacije i robe

komentara na www-u [3]. Prije nekoliko godina engle-

mogu slobodno kolati.

ski je moda bio lingua anca www-a jer je veina sadraja na www-u tada bila na engleskome, meutim, da-

2.1 JEZINE GRANICE KOE

EUROPSKO INFORMACIJSKO
DRUTVO

nas su se prilike u mnogome promijenile. Moglo bi se

rei kako je koliina sadraja na drugim jezicima (osobito azijskim i na arapskome) upravo eksplodirala. Iznenaujue je kako ovaj sveprisutni digitalni jaz prouzrokovan jezinim preprekama jo uvijek nije privukao do-

Ne moemo tono predvidjeti kako e izgledati budue

voljno pozornosti u javnim raspravama; pa ipak, upravo

informacijsko drutvo, ali s velikom se vjerojatnou

nas on navodi na gorue pitanje: Koji e europski jezici

moe oekivati kako e revolucija u komunikacijskim

napredovati i odrati se u umreenome informacijskome

tehnologijama na nove naine zbliiti ljude koji govore

drutvu i drutvu znanja, a koji e biti osueni na izumi-

razliite jezike. To e kod pojedinaca rezultirati potre-

ranje?

bom za uenjem novih jezika, a kod razvijatelja aplikacija potrebom za stvaranjem novih tehnolokih aplikacija, ne bi li se osiguralo uzajamno razumijevanje i omo-

2.2 OPASNOST ZA NAE JEZIKE

guio pristup razmjenjivome znanju. U globalnome

Dok je otkrie tiska neizmjerno pridonijelo razmjeni

gospodarskom i informacijskome prostoru raste inte-

obavijesti u Europi, ono je istodobno dovelo do izumi-

rakcija izmeu razliitih jezika, govornika i sadraja koja

ranja mnogih europskih jezika. Kako se na regionalnim

se odvija zahvaljujui novim vrstama medija. Trenu-

i manjinskim jezicima tiskalo rijetko, mnogi su jezici,

tana popularnost drutvenih mrea (kao to su Wiki-

npr. cornwallski ili dalmatski, bili ogranieni samo na

pedia, Facebook, Twitter, YouTube i od nedavno Go-

govorni oblik komunikacije to je ograniilo doseg nji-

ogle+) predstavlja samo vrak ledene sante.

hove uporabe. Hoe li Internet imati isti utjecaj na nae

dananje jezike? Osamdesetak europskih jezika najbogatiji je i najvanijih dio njezina kulturnoga nasljea i

Globalizacija gospodarstva i informacijskoga

prostora suoava nas sa sve vie razliitih
jezika, govornika i sadraja.

neizostavni dio jedinstvenoga drutvenoga modela [4].

Dok e iroko koriteni jezici kao engleski ili panjolski zacijelo odrati svoju prisutnost na rastuem tritu
digitalnoga drutva, mnogi bi europski jezici mogli biti

Danas bez ikakvih prepreka moemo u nekoliko se-

iskljueni iz digitalnih komunikacijskih kanala i postati

kunda na drugu stranu svijeta prebaciti gigabajte tek-

nevani za takvo umreeno drutvo. Time bi se s jedne

sta prije nego to uope shvatimo kako je on na jeziku

strane oslabio globalni poloaj Europe, a s druge strane,

koji uope ne razumijemo. Prema nedavnome izvjeu

takav bi razvoj bio u suprotnosti sa stratekim ciljem

Europske komisije, 57% internetskih korisnika u Europi

osiguravanja jednakoga sudjelovanja svakoga graanina

putem mree kupuje robu i usluge na jeziku koji nije nji-

EU bez obzira na njegov jezik.

hov vlastiti. Engleski je najei strani jezik, a slijede ga

Prema izvjeu UNESCO-a o viejezinosti jezici su

francuski, njemaki i panjolski. 55% korisnika ita sa-

kljuni medij za ostvarivanje temeljnih ljudskih prava

draje na stranome jeziku dok ih se samo 35% koristi

kao to su iskazivanje politikoga stava, obrazovanje i su-

stranim jezikom za pisanje poruka e-pote ili ostavljanje

djelovanje u drutvu [5].

 prevodimo www-stranice uporabom usluge on-line

Znatna raznolikost jezika u Europi jedno

je od najvanijih kulturnih dobara i bitan
je dio europskoga uspjeha.

prevoenja.
Jezine tehnologije, o kojima se podrobnije govori u
ovoj bijeloj knjizi, ine sr buduih inovativnih aplikacija. Jezine su tehnologije uobiajena potporna tehnologija unutar vee aplikacije kao to su navigacijski sustav

2.3 JEZINE SU TEHNOLOGIJE

KLJUNE POTPORNE
TEHNOLOGIJE
U prethodnim razdobljima ulaganje u jezike usredotoivalo se na uenje jezika i prevoenje. Na primjer, prema

ili trailica. Ove bijele knjige prikazuju stanje osnovnih

postignua u jezinim tehnologijama za svaki pojedini
jezik.

Svi e europski jezici trebati jezine

tehnologije koje e biti dostupne i prihvatljive.

nekim procjenama europsko trite prevoenja, tumaenja, lokalizacije programske podrke i prevoenja www-

Jezine se tehnologije sastoje od niza osnovih aplika-

stranica vrijedilo je 8,4 milijarde eura, a oekivao se nje-

cija koje omoguuju uporabu i obradbu jezika i govora

gov rast od 10% godinje [6]. No ak i uz takve prog-

unutar sloenijih aplikacijskih sustava. Svrha je ovih

noze rasta postojei kapaciteti nisu dovoljni za zadovo-

META-NET-ovih bijelih knjiga prikazati koliko su te

ljenje potreba niti sadanjih, a kamoli buduih potreba.

osnovne potporne jezine tehnologije razvijene za svaki

Najuvjerljivije rjeenje koje bi osiguralo irinu i dubinu

od europskih jezika. Europa treba robusne i dostupne

uporabe jezika u sutranjoj Europi jest uporaba odgova-

jezine tehnologije za sve europske jezike. Kako bi odr-

rajuih tehnologija, upravo kao to rabimo razne tehno-

ala svoj poloaj globalnoga predvonika u inovacijama,

logije pri rjeavanju npr. svojih transportnih ili energet-

Europa treba jezine tehnologije prilagoene svakome

skih potreba.

od svojih jezika, a one moraju biti robusne i dostupne

Jezine tehnologije usmjerene na sve vrste pisanoga ili

ne bi li se to lake integrirale u ire aplikacijsko okruje.

govorenoga teksta, pomau ljudima u suradnji, obavlja-

Bez jezinih tehnologija uskoro vie ne emo moi pos-

nju poslova, razmjeni znanja i sudjelovanju u drutve-

tii stvarno interaktivno, multimedijsko i viejezino

nim i politikim raspravama neovisno o stupnju usvo-

korisniko iskustvo.

jenih jezinih ili raunalnih vjetina. One ve esto

djeluju skrivene unutar sloenih raunalnih sustava koji
nam pomau kad:
traimo obavijesti koritenjem internetskih trailica;
provjeravamo pravopis ili gramatiku u obradniku
teksta;
gledamo preporuke za proizvode u on-line duanima;
sluamo glasovne upute navigacijskoga sustava;

2.4 MOGUNOSTI JEZINIH

TEHNOLOGIJA
U svijetu tiska tehnoloki je proboj predstavljalo brzo
umnoavanje slike teksta uporabom tiskarskoga stroja
pominim slovima. Meutim, istodbno je ljudima preputen teak posao traenja, pristupa, prevoenja i saimanja znanja irenoga i prenoenoga tako umnoenim
tekstovima. Morali smo ekati do Edisona koji je otkrio

kako zabiljeiti govor, ali ponovno je njegova tehnolo-

i miljenima sudionika, otkrivati emocionalne anitete,

gija stvarala analogne preslike.

uoavati krenje autorskih prava ili pratiti zloporabu.

Jezine nam tehnologije danas omoguuju pojednostav-

Jezine tehnologije Europskoj uniji pruaju upravo ne-

njivanje i automatizaciju postupaka kao to su strojno

sagledive ekonomski i kulturno znaajne mogunosti.

prevoenje, stvaranje sadraja, i upravljanje znanjem

One mogu pomoi u problemima koje donosi vieje-

na svim europskim jezicima. Jezine tehnologije tako-

zinost u Europi s obzirom da razliiti jezici prirodno

er stoje u pozadini intuitivnih govornih suelja za ku-

supostoje u europskom poslovanju, ustanovama i ko-

ansku elektroniku, strojeve, vozila, raunala i robote.

lama. Graani ele komunicirati onkraj jezinih granica

Premda ve postoje mnogi prototipovi, komercijalne i

koje jo uvijek postoje na europskome zajednikome tr-

industrijske primjene su jo uvijek u ranim stupnjevima

itu, a upravo bi jezine tehnologije mogle pomoi u

razvoja. Meutim, neka su nedavna postignua u is-

nadilaenju tih preostalih prepreka uz potpomaganje

traivanjima i razvoju otvorila jedinstvene mogunosti.

slobodne i otvorene uporabe bilo kojega jezika. Na-

Na primjer, strojnim prevoenjem ve se mogu dobiti

dalje, inovativne, viejezine jezine tehnologije nama

prijevodi prihvatljive tonosti unutar posebnih podru-

bi Europljanima takoer pomogle u komunikaciji s na-

ja, dok istodobno neke eksperimentalne aplikacije ve

im globalnim partnerima, a njima bi pomogle pri ra-

omoguuju dohvat viejezinih obavijesti i upravljanje

zvoju jezinih tehnologija u njihovim viejezinim za-

znanjem, kao i proizvodnju sadraja istodobno na mno-

jednicama. Jezine tehnologije postaju svojevrsne pot-

gim europskim jezicima.

porne tehnologije koje omoguuju nadii prepreke

jezine raznolikosti i ine razliite jezine zajednice me-

Viejezinost je pravilo, a ne iznimka.

usobno pristupanijima. Konano, jedno od aktivnijih

podruja istraivanja jest uporaba jezinih tehnologija
u spasilakim operacijama u unesreenim podrujima.

Kao to je to bio sluaj i s mnogim drugim tehnologi-

U takvim okrujima visoke opasnosti tonost prijevoda

jama, prve su jezine aplikacije, kao to su govorna ko-

moe znaiti razliku izmeu ivota i smrti: u budu-

risnika suelja i razgovorni sustavi, ponajprije razvijene

nosti e inteligentni roboti s viejezinim sposobnos-

u visokospecijaliziranim podrujima uporabe, ali neri-

tima moi ljudske spaavati ivote.

jetko uz ogranienu kakvou. Pa ipak i za takve aplikacije postoje ogromne trine mogunosti u obrazovanju
i zabavnoj industriji s ukljuivanjem jezinih tehnologija u raunalne igre, obrazovne sustave, knjinice, simulacijske sustave i sustave za uvjebavanje. Mobilne
obavijesne usluge, strojno potpomognuti programi ue-

2.5 IZAZOVI KOJI STOJE PRED

JEZINIM TEHNOLOGIJAMA

nja jezika, okruja za e-uenje, alati za samoprocjenu

Premda su u nekoliko proteklih godina jezine tehnolo-

i sustavi za pronalaenje plagijata samo su jo neki od

gije napravile znatan napredak, trenutaan je tempo teh-

primjera gdje jezine tehnologije igraju znaajnu ulogu.

nolokoga napretka i stvaranja novih proizvoda prespor.

Popularnost drutvenih mrea kao to su Twitter ili Fa-

Jezine tehnologije, koje su ve u irokoj uporabi, kao

cebook nagovijetaju dodatne potrebe za razraenim je-

to su provjernici pravopisa ili gramatike u obradnicima

zinih tehnologijama koje bi mogle nadgledati poruke,

teksta, uobiajeno su jednojezine, a dostupne su samo

saimati rasprave, predlagati opa kretanja u stavovima

za ogranien broj jezika.

Usluge on-line strojnoga prevoenja, premda korisne za

terijala koji opisuju jezino znanje u obliku apstraktnih

stvaranje opega dojma o emu je u nekom dokumentu

pravila, tablica i primjera.

rije, bore se s mnogim potekoama kad su nam potrebni visokokvalitetni i potpuni prijevodi. Zahvaljujui sloenosti prirodnih jezika, njihovo modeliranje u
obliku raunalnih programa i provjera u stvarnome i-

Ljudi usvajaju jezinu sposobnost na

dva razliita naina: uei na primjerima
i uei temeljna jezinih pravila.

votu, dugotrajan je i skup posao koji zahtijeva stalnu nancijsku potporu. Europa mora zadrati svoju vodeu

Kod jezinotehnolokih sustava dvije su osnovne vrste

ulogu u sueljavanju s tehnolokim izazovima viejezi-

usvajanja jezine sposob nosti, na slian nain kao i kod

noga drutva otkrivanjem novih naina za ubrzavanje

ljudi. Statistiki (ili podatkovno utemeljeni) pristupi

razvoja na tom podruju. To moe ukljuiti i razno-

stjeu jezino znanje iz golemih zbirki pojedinanih tek-

rodne pristupe kao to su napredak u raunarstvu, ali i

stnih primjera. Dok je dovoljno koristiti tekst na jed-

tehnike distribuirane ljudske potpore.

nome jeziku za treniranje npr. pravopisnoga provjernika, usporedni tekstovi na dva (ili vie) jezika potrebni
su za treniranje strojnoprevoditeljskih sustava. Algorit-

Tehnoloki se napredak mora ubrzati.

mima strojnoga uenja prepoznaju se obrasci kako se pojedine rijei, kratke fraze ili itave reenice prevode s jednoga jezika na drugi.
Meutim, za takve statistike pristupe potrebni su mili-

2.6 USVAJANJE JEZIKA KOD

LJUDI I STROJEVA

juni usporednih reenica za poveenje kakvoe izvedbe

Kako bismo prikazali na koji se nain raunala nose s

Provjera pravopisa u obradnicima teksta i usluge kao to

prirodnim jezikom i zato je iznimno teak zadatak pro-

su Google Search ili Google Translate poivaju u cije-

gramirati ih za obradbu razliitih jezika, pogledajmo na

losti na statistikim pristupima. Prednost statistikih

kratko kako ljudi usvajaju svoj prvi i ostale jezike, a po-

sustava je to strojevi ue brzo i u kontinuiranim ciklu-

tom pogledajmo kako djeluju jezinotehnoloki sustavi.

sima treniranja premda kakvoa takvih sustava nerijetko

Ljudi usvajaju jezine sposobnosti na dva razliita na-

oscilira.

ina. Mala djeca usvajaju jezik sluanjem i praenjem in-

Sustavi temeljeni na pravilima predstavljaju drugu os-

terakcija izmeu svojih roditelja, brae i sestara te ostalih

novnu vrstu jezinih tehnologija, a time i sustava za

u obitelji. U otprilike dvogodinjoj dobi sami poinju

strojno prevoenje. Strunjaci s podruja jezikoslovlja,

proizvoditi prve rijei i kratke fraze. To je mogue samo

raunalnoga jezikoslovlja i raunarstva moraju kodirati

zato jer ljudski rod ve ima genetsku predispoziciju za

gramatike analise (prijevodna pravila) i sastaviti popise

imitiranje i racionalizaciju onoga to uju.

rijei (rjenike). Izgradnja sustava temeljenoga na pra-

Uenje drugoga jezika u kasnijoj dobi obino trai vie

vilima iznimno je vremenski i poslovno zahtjevna, a ne

kognitivnoga napora ukoliko dijete nije uronjeno u je-

moe se provesti bez visokospecijaliziranih strunjaka.

zinu zajednicu izvornih govornika. U kolskoj se dobi

Neki od vodeih strojnoprevoditeljskih sustava temelje-

strani jezici obino usvajaju uenjem njihove gramatike

nih na pravilima u stalnom su razvoju ve dvadesetak go-

strukture, rjenika i pravopisa iz knjiga i obrazovnih ma-

dina. Prednost sustava temeljenih na pravilima je u tome

takvih sustava. To je jedan od razloga zato sastavljai

trailica tee skupiti to je vie mogue pisanoga teksta.

to strunjaci imaju mogunost istananijega upravljanja obradbom jezika. Zbog toga je mogue sustavno ispravljati pogrjeke u programskoj podrci i pruati ko-

Dvije osnovne vrste jezinih tehnologija

usvajaju jezino znanje na slian nain.

risniku podrobne povratne obavijesti, osobito kad se na

pravilima temeljeni sustavi koriste za uenje jezika. Na

Kao to smo vidjeli u ovome poglavlju, mnoge aplika-

alost, zbog svoje visoke cijene, jezine tehnologije te-

cije, koje se svakodnevno na iroko koriste u dananjem

meljene na pravilima isplative su samo za jezike s velikim

informacijskome drutvu, u mnogome ovise o jezinih

brojem govornika.

tehnologijama, osobito europskome gospodarskom i in-

Kako su prednosti i nedostatci statistikih pristupa i

formacijskome prostoru. Premda su te tehnologije pos-

pristupa temeljenih na pravilima meusobno nadopu-

tigle znatan napredak u nekoliko proteklih godina, jo

njui, trenutana se istraivanja usredotouju na hi-

uvijek postoje ogomne mogunosti za poboljanje kak-

bridne pristupe koji kombiniraju ova dva pristupa. Me-

voe jezinotehnolokih sustava. U sljedeem emo po-

utim, hibridni sustavi do sad su bili znatno manje us-

glavlju opisati ulogu hrvatskoga jezika u europskome in-

pjeni u industrijskim uvjetima nego u istraivakim la-

formacijskome drutvu i procijeniti trenutano stanje

boratorijima.

jezinih tehnologija za hrvatski jezik.

3
HRVATSKI JEZIK U EUROPSKOME
INFORMACIJSKOME DRUTVU
3.1 OPE INJENICE

Ustava: U Republici Hrvatskoj u slubenoj je uporabi

Hrvatski jezik pripada zapadnoj junoslavenskoj pod-

jedinicama uz hrvatski jezik i latinino pismo u slu-

skupini slavenske grane inoeuropske jezine porodice.

benu se uporabu moe uvesti i drugi jezik te irilino ili

U ovome trenutku hrvatski jezik broji preko 5,5 mili-

koje drugo pismo pod uvjetima propisanima zakonom.

juna izvornih govornika. Hrvatski se jezik sastoji od na-

Kako se 2013. oekuje pristup Hrvatske Europskoj uniji,

rjeja i nacionalnoga standardnoga jezika Hrvata, koji

hrvatski e jezik postati 24. slubeni jezik EU-a.

hrvatski jezik i latinino pismo. U pojedinim lokalnim

je slubeni jezik vie od 4 milijuna stanovnika Republike Hrvatske, a uz bonjaki i srpski takoer je jedan
od tri slubena jezika Bosne i Hercegovine gdje ga govori oko 700.000 govornika. Takoer, hrvatskim jezikom govore i mnogi pripadnici nacionalnih manjina u

Hrvatskim se jezikom slui vlada i administracija,

sve razine obrazovanja, a i jezik je na kojem se
odvija poslovanje i svakodnevna komunikacija u
Republici Hrvatskoj.

Hrvatskoj kao i autohtone hrvatske etnike i jezine manjine u Srbiji, Crnoj Gori, Sloveniji, Madarskoj, Aus-

U Hrvatskoj jo uvijek ne postoji jedinstven jezini za-

triji, Slovakoj i Italiji, koje ili obitavaju na teritorijima

kon koji bi regulirao slubenu uporabu jezika u jav-

nekadanjih hrvatskih zemalja ili su iselili tijekom sto-

nosti. Uvoenje zakona o jeziku pokuano je u neko-

ljea u povijesno uvjetovanim selidbama. Zbog znatne

liko navrata od stjecanja hrvatske neovisnosti, ali niti u

gospodarski i politiki uvjetovane emigracije u 20. sto-

jednom sluaju nije dobivena dovoljna potpora hrvat-

ljeu i nakon dvaju svjetskih ratova, hrvatski se takoer

ske Vlade te niti jedan prijedlog nije uao u saborsku

govori u mnogobrojnim hrvatskim zajednicama u cije-

proceduru. Zadnji se takav pokuaj dogodio u travnju

lome nizu europskih i prekomorskih zemalja. Najvee

2010. Meutim, u zakonima o obrazovanju, sudskim

hrvatsko gospodarsko iseljenitvo smjeteno je u Nje-

postupcima itd. postoje lanci koji reguliraju uporabu

makoj, potom u SAD, Kanadi i Australiji. Aktivna

hrvatskoga kao slubenoga dravnoga jezika. Do sada,

uporaba hrvatskoga uglavnom ovisi o iseljenikome na-

zakonodavstvo ne zahtijeva obvezatnu provjeru ili ispi-

rataju kojem govornici pripadaju. Pa ipak, u mnogim

tivanje znanja hrvatskoga jezika kao uvjet za naturaliza-

zemljama, osobito europskima, postoje dodatni kolski

ciju. Zakon o hrvatskom dravljanstvu [7] prepostavlja

programi programi na hrvatskome koje organizira i po-

da stranac, koji trai stjecanje hrvatskoga dravljanstva,

dupire hrvatska Vlada.

poznaje hrvatski jezik i pismo.

Slubeni je status hrvatskoga jezika u Hrvatskoj odre-

Prema popisu stanovnitva iz 2001. Hrvatska je imala

en Ustavom Republike Hrvatske. Prema lanku 12

4.437.460 stanovnika od kojih su 89,63% Hrvati. Srbi

su najzastupljenija nacionalna manjina s 4,54% stanov-

u Austriji aktivno slue gradianskim hrvatskim. Ova

nitva dok svaka od preostalih nacionalnih manjina za-

je varijanta hrvatskoga jezika, standardizirana u skladu

uzima manje od 0,5% stanovnitva: Bonjaci (0,47%),

s poneto drukijim naelima od standardnoga hrvat-

Albanci (0,34%), Slovenci (0,30%), Crnogorci (0,11%)

skoga, jedan od austrijskih slubenih manjinskih jezika.

i ostali u jo manjim postotcima. Hrvatski je jezik

itav je niz djejih vrtia i kola u Gradiu u kojima se

materinski jezik za 96% stanovnika. Nacionalne ma-

rabi gradianski hrvatski. S druge strane, hrvatski stan-

njine izjasnile su se kako govore sljedee jezike: alban-

dardni jezik slubeni je manjinski jezik u Madarskoj. U

ski, bonjaki, bugarski, eki, hebrejski, madarski, nje-

Italiji trenutano ivi oko 3.000 Hrvata koji se slue va-

maki, istrorumunjski, talijanski, makedonski, crnogor-

rijantom hrvatskoga zvanom moliki hrvatski i on se ta-

ski, poljski, romski, rumunjski, ruski, rusinski, slovaki,

koer ui u kolama u tri opine nastanjene Hrvatima u

srpski, turski i ukrajinski. Jezici etiri manjine, srpski,

Moliseu. Broj Hrvata u Srbiji, posebice u pokrajini Voj-

madarski, talijanski i eki, stekli su status jezika i pisma

vodini gdje su Hrvati priznata nacionalna manjina, te-

u slubenoj uporabi u odreenim podrujima prema

ko je tono utvrditi jer se dio etnikih Hrvata izjanjava

udjelu njihovih govornika u ukupnome stanovnitvu

kao tzv. Bunjevci uglavnom zbog politikih razloga.

koji mora iznositi barem 1/3 svih stanovnika na podru-

Premda je mnogo Hrvata izgnano iz Srbije nakon to

ju lokalne samoupravi. Od 2009. u Hrvatskoj postoji

je Hrvatska stekla svoju neovisnost od Jugoslavije, pret-

27 podruja gdje nacionalne manjine imaju pravo slu-

postavlja se kako u Srbiji jo uvijek ivi vie od 100.000

bene uporabe vlastitoga jezika u lokalnoj administraciji.

Hrvata. U ostalim europskim zemljama hrvatska autoh-

To se pravo u visokome omjeru primjenjuje u Istarskoj

tona manjina ivi u Crnoj Gori (7.000 do 10.000), e-

upaniji gdje je talijanski materinski jezik 20.521 sta-

koj (manje od 1.000), Slovakoj (4.000) i Rumunjskoj

novnika, ali su dvojezini cestovni natpisi prisutni i u

(7.500). Broj Hrvata u Sloveniji je oko 50.000, ali samo

dijelovima gdje nema talijanske manjine. Republika je

je malen broj njih stvarna autohtona manjina, ponaj-

Hrvatska raticirala Europsku povelju o regionalnim i

prije u naseljima uz granicu, a veina ih predstavlja ne-

manjinskim jezicima 1997.

davno gospodarsko iseljenitvo. Hrvatski ima status ma-

Jo nije objavljena slubena statistika o jezinoj uporabi

prikupljena nedavno provedenim popisom iz 2011. usklaenim s meunarodnim statistikim normama, koji

njinskoga jezika u Srbiji (kao jedan od sedam slubenih

jezika pokrajine Vojvodine), Crnoj Gori, Austriji, Madarskoj i Italiji.

je tako obuhvatio sve dravljane Republike Hrvatske,

strane dravljane i apatride koji borave u Republici Hrvatskoj.

3.2 HRVATSKA NARJEJA

Slika hrvatskih narjeja sastavljena je od tri narjene sku-

Hrvatska ima brojno iseljenitvo koje esto jo uvijek go-

pine: akavske, kajkavske i tokavske (v. sliku 2). Mjesni

vori hrvatski jezik (v. sliku 1). Hrvatske etnike i jezine

govori, koji pripadaju nekom od triju narjeja, govore se

manjine ive u mnogim europskim zemljama kao pos-

po cijeloj Republici Hrvatskoj. Sva hrvatska narjeja pri-

ljedica povijesnih selidaba, zapoetih jo u 16. stoljeu,

padaju srednjo-junoslavenskome dijasistemu slavenske

kao nedavnih, mahom gospodarski i politiki uvjetova-

jezine grane i u junoslavenskom prostoru obuhvaa

nih. Najbrojnije skupine su tzv. Gradianski Hrvati u

dio dijalekatnoga kontinuuma izmeu slovenskoga tipa

Austriji (pretpostavlja se oko 50.000), a otprilike slian

na sjeverozapadu i makedonsko-bugarskoga tipa na ju-

broj Hrvata ivi u Madarskoj. Gradianski se Hrvati

goistoku. Imena triju narjeja izvedena su iz oblika

10

1: Hrvati u susjednim dravama [8]

11

upitne zamjenice a, kaj i to (lat. quid). Meutim, na

junoslavenskome prostoru ta je klasikacija relevantna
samo za hrvatske dijalekte i rezultat je potreba hrvatske

3.3 STANDARDIZACIJA
HRVATSKOGA JEZIKA

jezine zajednice. Slovenci koriste zamjenicu kaj, ali slo-

Tisuljetna povijest hrvatskoga jezika potvrena je tek-

venski jezik ne pripada u kajkavsko narjeje. Bonjaci,

stovima pisanim jo krajem 10. stoljea ili poetkom

Crnogorci, Srbi kao i Bugari, Makedonci i svi istoni

11. stoljea, u vrijeme kad su se tri hrvatska narjeja (a-

Slaveni koriste to, ali njihovi jezici ne pripadaju u to-

kavski, tokavski i kajkavski) poela oblikovati. Sva su tri

kavsko narjeje u istome smislu u kojem je tokavski hr-

hrvatska narjeja odigrala vanu ulogu u stvaranju hrvat-

vatsko narjeje. Srbi, Crnogorci i Bonjaci nemaju oblik

skoga knjievnoga jezika (razliitih narjenih osnovica)

te upitne zamjenice kao kriterij za razlikovanje svojih

i oblikovanju hrvatske jezine kulture koja je dovela do

narjeja. Kad je rije o tokavskome, arhaini akav-

standardnoga hrvatskoga jezika izgraenoga na tokav-

ski (tzv. slavonski) govore samo Hrvati, novotokavski

skoj osnovici.

ikavski i ijekavsko-akavski govore Hrvati i Bonjaci, a

novotokavski ijekavski govore Hrvati iz irega dubrovakoga podruja, ali takoer i ostali juni Slaveni. Hrvati u Gradiu (Austrija, Madarska, Slovaka) mahom

Jeste li znali da je etimologija rijei kravata

dolazi od Croate i da se iz francuskoga
proirila na ostale jezike u 17. stoljeu?

govore akavski, a rijetko tokavski ili kajkavski. Hrvati

u talijanskoj pokrajini Molise govore arhainim tokav-

Prvi jasni pokuaj oblikovanja hrvatskoga standardnoga

skim dok Hrvati u Karaevu, Rumunjska govore torla-

jezika pojavio se u 17. stoljeu kad je veina hrvatske et-

kim narjejem.

nike zajednice osobito nakon gramatike i drugih djela

Bartola Kaia (1575-1650) i rascvjetale renesansne i ba-

Uslijed mnogobrojnih, esto prisilnih iseljavanja, pros-

rokne knjievnosti tokavskoga Dubrovnika prepoz-

torna se raspodjela pojedinih hrvatskih narjeja stubo-

nala jezinu strukturu tokavskoga (isprva s ikavskim re-

kom promijenila od srednjega vijeka. akavski i kajkav-

eksom jata, ali kasnije s jekavskim) kao najbolje polazi-

ski su u prolosti bili rasporeeni na znatno irem po-

te za sastavljanje nadregionalnoga hrvatskoga knjiev-

druju, ali danas prevladava tokavsko narjeje. Prije is-

noga jezika. Unato odabiru jedne jezine osnovice za

eljavanja akavsko se narjeje rabilo na sjeveru do rijeka

sastavljanje svoga standardnoga jezika, Hrvati nisu od-

Kupe i Save a na istoku do crte Una-Dinara-Cetina. Na-

bacili postignua viestoljetne jezine kulture razliitih

kon migracija akavsko je narjeje ogranieno na obalno

narjenih osnovica unutar hrvatskoga knjievnoga je-

podruje i otoke, dok su se akavski govori u unutra-

zika (kajkavsko-tokavsko-akavski hibrid) koji je obi-

njosti poeli razlikovati prema koliini tokavskoga utje-

ljeio i povijest hrvatske etnike zajednice. Premda je

caja. Kajkavsko se narjeje takoer nekad prostiralo is-

standardizacija jezika Hrvata temeljena na tokavskome

tonije gdje danas prevladava tokavsko.

narjeju zapoela vrlo rano, narodno se jezino jedinstvo postiglo tek u vrijeme Ilirskoga narodnoga prepo-

akavsko, kajkavsko i tokavsko narjeje razlikuju se na

roda (poevi od 1835.) kad je mala skupina Hrvata,

svim jezinim razinama: fonolokoj, morfolokoj, sin-

koji su se do tada sluili kajkavskim idiomom, takoer

taktikoj i leksikoj i svaka od tih razina ukljuuje brojne

prihvatili tokavski hrvatski standardni jezik. Tijekom

arhaizme, ali i inovacije karakteristine za odreeno na-

veine 20. stoljea hrvatski se standardni jezik razvijao

rjeje.

u razliitim junoslavenskim dravnim jedinicama pod

12

2: Zemljovid narjeja u Republici Hrvatskoj

13

razliitim imenima, a bio je predstavljan kao varijanta

velikim gradskim sreditima koja su smjetena izvan no-

tzv. hrvatsko-srpskoga (srpsko-hrvatskoga) jezika, po-

votokavskoga podruja (npr. kontinuitt / kontinutt).

najprije iz politikih razloga. To je naputeno s demo-

Naglasak i duina mogu se povremeno rabiti za razli-

kratskim drutveno-politikim promjenama 1990.

kovanje znaenja izmeu leksikih jedinica ili njihovih

Razliite stilizacije hrvatskoga jezika oblikovane su jo

oblika, npr. gr d : grd, n (gen. jd.) : ne (nom. mn.).

davno u iseljenitvu (npr. gradianski hrvatski, moliki

U hrvatskome neke rijei nemaju vlastiti naglasak (na-

hrvatski). Hrvatska je pisana kultura obiljeena upora-

slonjenice) ve u naglasnoj rijei prednaglasnice mogu

bom triju pisama (glagoljica, irilica, latinica) meu ko-

preuzeti naglasak prenesen s naglaene rijei ukoliko je

jima je latinica meu Hrvatima prevladava od 16. sto-

naglasak silazni i na prvom je slogu (grd : grd). Kod

ljea. Njezina uporaba nije bila normirana niti usustav-

zanaglasnica to nije mogue. Prenoenje naglaska na

ljena sve do 1835. kad je Ljudevit Gaj dao hrvatskoj la-

prednaglasnicu postaje sve rjee, osobito u gradskim sre-

tinici dananji oblik.

ditima izvan neotokavskoga podruja.

U hrvatskome se standardnome jeziku nalaze mnoge fo-

3.4 OSOBINE HRVATSKOGA

JEZIKA

noloki (nom. jd. sladak : gen. jd. slatkoga, nom. jd. dio :
gen. jd. dijela) i morfonoloki uvjetovane promjene
(nom. jd. majka : dat. jd. majci, nom. jd. junak :

3.4.1 Fonetika, fonologija, morfonologija

vok. jd. junae).

Fonemski inventar hrvatskoga standardnoga jezika sas-

esto je u govoru pod utjecajem lokalnoga narjeja,

toji se od 5 samoglasnika (a, e, i, o, u) i 25 suglasnika (m,

npr. na akavskome Kvarneru prevladava zatvorno t

v, n, l, r, j, nj, lj, p, b, f, s, z, c, t, d, , , , , , d, h, k, g).

umjesto bezvunoga poluzatvornoga , ili u sjeveroza-

Akustine i artikulacijske osobine samoglasnika ne mije-

padnome kajkavskome znakovito je nerazlikovanje iz-

njaju se s obzirom na mjesto izgovora (bez obzira nalazi

meu ili d.

Regionalna primjena hrvatskoga standardnoga jezika

li se u kratkom, dugom, naglaenom ili nenaglaenom

slogu). Uz tih 5 samoglasnika postoji i samoglasniko r
(crn niger) i dvoglas ie, koji se u pismu biljei kao je/ije

3.4.2 Morfologija

(djelo, odijelo).

Hrvatski standardni jezik razlikuje 10 vrsta rijei od ko-

Naglasni sustav sastoji se od 4 naglaska (dva duga na-

jih su pet promjenljive (imenice, pridjevi, zamjenice,

glaska: s uzlaznim i silaznim tonom i dva kratka na-

brojevi, glagoli), etiri nepromjenljive (prijedlozi, vez-

glaska: s uzlaznim i silaznim tonom) i zanaglasne du-

nici, uzvici, estice), a prilozi su promjenljivi samo u

ine. Naglasni je sustav standardnoga hrvatskoga jezika

komparaciji.

novotokavski premda danas postoje mnoga odstupa-

Gramatike kategorije koje se nalaze u veine promjen-

nja od naglasnih modela kodiciranih u drugoj polo-

ljivih rijei jesu rod (tri vrijednosti: muki, enski, sred-

vici 19. stoljea. Mjesto naglaska nije vezano uz poje-

nji), broj (dvije vrijednosti: jednina, mnoina), pade

dini slog, nego raspodjela naglasaka podlijee stanovi-

(sedam vrijednosti: nominativ, genitiv, dativ, akuza-

tim ogranienjima (npr. zadnji slog vieslone rijei u

tiv, vokativ, lokativ, instrumental). Neke sklonjive ri-

naelu ne moe biti naglaen, silazni naglasci ostvaruju

jei imaju i neke posebne kategorije (npr. odreenost se

se samo na prvome slogu rijei koje nisu sloenice, itd.)

obiljeava na pridjevima zasebnim nizom ektivnih nas-

Ova se pravila kre u svakodnevnome govoru, osobito u

tavaka; ivo/neivo se obiljeava odabirom nastavka za

14

akuzativ jednine imenica mukoga roda; imenice mogu

jetla (npr. iz kajkavskoga kukac, hlae, rjenik, ili iz a-

biti konkretne, tvarne, kategorijalne ili zbirne; itd.). Ko-

kavskoga spuva). Pored toga, cjelina hrvatskoga jezika

njugirane rijei (glagoli) obiljeene su kategorijama: na-

bila je stalno izloena izravnim ili neizravnim dodirima

ina (etiri vrijednosti: indikativ, imperativ, kondici-

s drugim jezicima i kulturama. Hrvatski se jezik istie iz-

onal, optativ), lica (tri vrijednosti: prvo, drugo, tree),

meu ostalih junoslavenskih jezika znatnim leksikim

broja (dvije vrijednosti: jednina, mnoina), stanja (dvije

utjecajima pristiglim iz romanskih jezika (supstratni tra-

vrijednosti: aktiv, pasiv) i vremena (sedam vrijednosti:

govi dalmatskoga jezika jesu npr. jarbol, tunj). Talijanski

prezent, imperfekt, aorist, perfekt, pluskvamperfekt, fu-

je jezik bio utjecajan u priobalju (osobito u dijelovima

tur I., futur II.). Glagoli biti (esse) and htjeti (volere) u

pod negdanjom mletakom dominacijom), a njemaki

hrvatskome su pomoni glagoli. Glagoli takoer posje-

i do neke mjere madarski, u kontinentalnoj Hrvatskoj.

duju sloen sustav glagolskih vidova (svreni i nesvreni s

dodatnim podvrijednostima kao to su poetni, uestali

Crkvenoslavenski je knjievni jezik ostavio tragove u sta-

itd.), a mogu ukljuivati i osobinu prijelaznosti. Pridjevi

rijim razdobljima hrvatskoga jezika, ali nije imao zna-

i prilozi mogu se pojaviti i u kompariranim oblicima (tri

ajnijega utjecaja tijekom razdoblja u kojem se obliko-

vrijednosti: pozitiv, komparativ, superlativ).

vao standardni jezik. Ruski jezik nije ostavio tako du-

U hrvatskome postoje dvije osnovne vrste sklonidbe:

bokoga traga u hrvatskome kao to je to uinio u susjed-

imenina sklonidba (imenice i neodreeni oblici pri-

nome srpskome standardnome jeziku. Utjecaj vokabu-

djeva) i zamjenino-pri-djevska sklonidba (zamjenice,

lara klasinih jezika (latinskoga i grkoga) sveprisutan je

odreeni oblici pridjeva, brojevi). Svaki imenini rod

u hrvatskoj kulturi, a ponajprije u intelektualnom vo-

ima svoju sklonidbu (a-vrsta za muki i srednji rod, e vr-

kabularu i znanstvenome nazivlju. Tijekom razdoblja

sta za enski), a postoji i posebna i-vrsta (imenice en-

srednjohrvatskoga jezika (16.-18. stoljee) intenzivno su

skoga roda). Imenina sklonidba prikazana je na slici 3.

u hrvatski ulazile posuenice iz turskoga, osobito rijei

Nastavci za zamjenino-pridjevsku sklonidbu prikazani

za predmete iz svakodnevnoga ivota. Vano je napo-

su na slici 4.

menuti kako zbog ranoga iseljavanja, u gradianskome

Rijei se u hrvatskome tvore derivacijom i slaganjem.

hrvatskome nema turskih posuenica, pa ak niti onih

Postoji nekoliko razliitih naina tvorbe rijei: su-

koje se u standardnome hrvatskome vie niti ne osje-

ksalna (star-ac), preksalno suksalna (do-iot-an),

aju stranim rijeima (npr. bubreg, izma, jastuk itd.).

nesuksalno slaganje (plaidrug), suksalno slaganje

Umjesto tih rijei u gradianskome hrvatskome rabe se

(vanjskopolitiki), srastanje (uz-brdo), slaganje pokrata

starije hrvatske rijei zajednikoga slavenskoga podrije-

(Varteks) i pretvorba (mlada). Najea je suksalna

tla, te je stoga on vrlo bitan za uvid u povijest hrvatskoga

tvorba.

leksikoga inventara. Njemaki i francuski nekad su takoer utjecali na hrvatski vokabular, a od druge polovice

3.4.3 Rjenik, frazeologija, nazivlje

20. stoljea utjecaj engleskoga jaa. eki, premda ne u

Temeljni se leksiki sloj hrvatskoga standardnoga jezika,

vokabular u nekoliko navrata, osobito tijekom 19. sto-

osim praslavenskoga leksikoga nasljea, sastoji od to-

ljea za vrijeme izgradnje strunoga nazivlja koju je bio

kavskoga vokabulara uz primjese vokabulara drugih hr-

izveo Bogoslav ulek (npr. asopis, kisik, duik, odik).

vatskih narjeja i vokabulara naslijeenoga iz knjiev-

Za vrijeme Jugoslavija, na hrvatski je utjecao i srpski, a

noga jezika raznih dijalekatnih stilizacija starijega podri-

osobitu je za to zaslugu imala federalna administracija.

izravnome kontaktu, imao je znaajan utjecaj na hrvatski

15

imenina sklonidba

nom. i gen. jd.

nom. mn.

a-vrsta muki rod

a-vrsta srednji rod
e-vrsta enski rod
i-vrsta enski rod

opis, opisa
sunce, sunca
ena, ene
no, noi

opisi
sunca
ene
noi

3: Imenina sklonidba u hrvatskome jeziku

pade

muki rod

srednji rod

enski rod

jednina
N
G
D
A
V
L
I

-i
-og(a) -eg(a)
-om(u/e) -em(u/e)
=N/=G
=N
-om(u/e) -em(u/e)
-im

-o -e
-og(a) -eg(a)
-om(u/e) -em(u/e)
=N
=N
-om(u/e) -em(u/e)
-im

-a
-e
-oj
-u
=N
=D
-om

mnoina
N
G
D
A
V
L
I

-i
-ih
-im(a)
-e
=N
=D
=D

-a
-ih
-im(a)
=N
=N
=D
=D

-e
-ih
-im(a)
=N
=N
=D
=D

4: Zamjeniko-pridjevska sklonidba u hrvatskome jeziku

16

Puristike tenje u vokabularu pojavljivale su se od vre-

vid, a glagolski oblici takoer izraavaju glagolsko vri-

mena do vremena od 16. do 20. stoljea (npr. Zorani,

jeme i modalna znaenja. Organizacija sloenih ree-

Ritter Vitezovi, Reljkovi, razdoblje 1941.-1945.).

nica moe biti nezavisna ili zavisna (uz prisutnost vez-

Kontinuitet od davnih vremena do suvremenoga hrvat-

nika ili bez njih). Novija je pojava u suvremenome je-

skoga standardnoga jezika i sudjelovanje triju narjeja u

ziku ogranienje uporabe zajednikoga slavenskoga ge-

izgradnji hrvatskoga standardnoga jezika moe se uoiti

nitiva (Nije olio vina), posvojne genitivne konstrukcije

u njegovoj razvijenoj i bogatoj frazeologiji (npr. u svojim

izbjegavaju se u korist posvojnih pridjeva (majina kua

umjetnikim tekstovima iz 16. stoljea Maruli rabi fra-

umjesto kua majke), a uporaba prolih vremena (imper-

zem zgubiti glas = biti postien, izgubiti lice, dok Zora-

fekt, aorist i pluskvamperfekt) je sve ogranienija. U su-

ni rabi frazem u magnutje oka = odmah, koji su gotovo

vremenome su hrvatskome pasivne konstrukcije znatno

isti kao frazemi izgubiti glas i u trenu oka u dananjemu

rjee nego u starijem hrvatskome.

hrvatskome standardnome jeziku temeljenome na tokavskoj osnovici).

Nazivlje u pojedinim strunim podrujima zapoelo se
razvijati ve u 16. stoljeu, a to je potvreno mnogobroj-

3.4.5 Pravopis

nim hrvatskim (ponajvie viejezinim) rjenicima sastavljenim od 16. do 20. stoljea. U 19. stoljeu njemaki
i eki imali su iznimno jak utjecaj na hrvatsko nazivlje,
a engleski je danas preuzeo tu ulogu.

3.4.4 Sintaksa

Premda je povijest hrvatske kulture obiljeena uporabom triju pisama (glagoljica, irilica, latinica), latinica
u Hrvata prevladava od 16. stoljea. Hrvatska latinina
abeceda nije bila u cijelosti standardizirana do 1835. kad
joj Ljudevit Gaj daje dananji oblik. Sastoji se od 30

Hrvatski jezik pripada skupini jezika obiljeenih SVO

slova, od koji su tri dvoslovi (d, lj, nj), a ostala su jed-

sintaktikom strukturom (Marija oli Ivana) i relativno

noslovi od ega pet s dijakritikim znacima (, , , , ).

slobodnim redom rijei (mnogobrojne permutacije sas-

U akademskim krugovima, osobito pri tiskanju tekstova

tavnica mogue su uz neka ogranienja kao to je smje-

hrvatske pismene batine, dvoslovi d, lj i nj se mogu za-

taj nenaglasnica). Glede informacijske strukture ree-

mijeniti s , and . Slova q, x, y, w ne postoje izvorno

nica, temeljno je pravilo u stilistiki neutralnome di-

u hrvatskoj abecedi premda se rabe za pisanje stranih

skursu da se na prvo mjesto smjeta tema (stara obavi-

imena. Hrvatska latinica dana je na slici 5.

jest), a slijedi u rema (nova obavijest, primjedba).

Hrvatski je pravopis fonoloko-morfonoloki jer pred-

Subjekt u reenici ne mora biti izrijekom naveden, a

stavlja stapanje dvaju pravopisnih naela: nadreenoga

njegovo je isputanje poeljno ukoliko bi ga se trebalo

fonolokoga (npr. biljeenje asimilacije) i podreenoga

ponavljati vie puta unutar neposredne okoline. Obve-

morfonolokoga (npr. podcrtati). Razmak izmeu ri-

zatna je dvostruka negacija (Nitko ga nije olio). Sro-

jei je logiki, a ne gramatiki (kakav je bio nekada). Za

nost sastavnica u rodu, broju i padeu je tipina za struk-

hrvatski je pravopis tipino da se pisanje stranih imena

turu hrvatskih reenica.

ne prilagouje izgovoru ili grafemskom sastavu hrvat-

U hrvatskome standardnome jeziku postoji sedam pa-

ske abecede, a i oblini se nastavci uklapaju u itavu ri-

dea, a oblici se mogu kombinirati s prijedlozima (obve-

je (npr. John, a ne Don; Washington, a ne Vaington;

zatni uz lokativ). Bitna odrednica hrvatskih glagola jest

Johna, a ne John-a).

17

velika slova
A
L

B
LJ

C
M

NJ

D
O

D
P

E
S

G
T

H
U

I
V

J
Z

e
s

g
t

h
u

i
v

j
z

mala slova
a
l

b
lj

c
m

nj

d
o

d
p

5: Hrvatska latinina abeceda

3.4.6 Onomastika
Hrvatska imena predstavljaju vane spomenike jezinoga, kulturnoga i drutvenoga nasljea ljudi koji su ih
napravili. Stoga i osobna imena (antroponimi) i imena
mjesta (toponimi) ine vaan dio hrvatske jezine kulture. Ozemlje dananje Hrvatske, u grubo ogranieno
rijekom Dravom na sjeveru, rijekom Dunavom na istoku i Jadranskim morem na jugu, vrlo se ilustrativno
reektira u bogatom raslojavanu zemljopisnih imena.

van je u skladu s praslavenskim imenskim obrascima koji

su pak slijedili zajednike indoeuropske obrasce oblikovanja imena. Patronimici jo uvijek ine najvei dio
inventara prezimena, ali za razliku od ruskoga, danas
vie nisu produktivni i ostaju neizmijenjeni kao zamrznuta prezimena koja su uklopljena u ektivni sustav
kao imenice. U suprotnosti s hrvatskim toponomastikim sustavom gdje gotovo da i nema turskoga utjecaja,
mnoga su hrvatska prezimena oblikovana iz turskih posuenica hrvatskim tvorbenim nastavcima. Tome je razlog injenica kako je veina prezimena u Hrvatskoj stvo-

Jeste li znali kako su Hrvati prvi slavenski

narod koji je uveo prezimena u 12. stoljeu?

rena nakon tridentinskoga koncila u 16. stoljeu, u vrijeme kad je velik dio hrvatskih zemalja bio pod turskom
vlau.

To obilno raslojavanje u hrvatskoj toponimiji odraz je

viestoljetnoga suivota razliitih etnikih skupina koje

esto su svjedocima najstarijih promjena u samome hr-

3.5 ODNOS HRVATSKOGA

STANDARDNOGA JEZIKA S
OSTALIM JEZICIMA TOKAVSKE
OSNOVICE

vatskome jeziku.

etiri nacionalna jezika, hrvatski, srpski i od nedavna,

Kako se hrvatski jezik razvijao preko vjerskih (pretkr-

bonjaki i crnogorski, svi dijele tokavsku strukturnu

anstvo i kranstvo), kulturnih i civilizacijskih granica,

osnovicu, meutim, tradicije i nadstrukture ovih je-

tragovi i Istoka i Zapada mogu se uoiti u hrvatskim ime-

zika su poprilino razliite. to razlikuje hrvatsku je-

nima. Kad je rije o imenima osoba, Hrvati su prvi sla-

zinu povijest i kulturu od ostalih junoslavenskih jezika

venski narod koji je uveo prezimena (od 12. stoljea) uz-

jest odnos izmeu svih triju narjeja (kajkavsko, akav-

du jadranske obale uslijed izravnoga romanskoga kul-

sko, tokavsko) koji odnos postojano obogauje hrvat-

turnoga utjecaja. Najstariji sloj hrvatskih imena obliko-

ski standardni jezik tokavske osnovice. Zbog razlii-

su nastanjivale istonu obalu Jadrana i njezino zalee u

povijesti. Stoljea jezinoga proimanja i stapanja razliitih kulturnih tradicija ostavila su neizbrisiv trag u hrvatskoj toponimiji. tovie, potvrena imena mjesta po-

18

tih polaznih uvjeta (nepostojanje osnovnoga, zajednikoga standarda) i razliitih tradicija u jezinome kultiviranju i standardizaciji, zbog razjedinjenja neotokav-

3.6 SKRB O JEZIKU U

HRVATSKOJ

skih struktura i razlika u jezinim nadstrukturama jedan

Vijee za normu hrvatskoga standardnoga jezika usta-

zajedniki monolitni standarni jezik nikad nije bio us-

novljeno je odlukom Ministarstva znanosti, obrazova-

pio biti oblikovan tijekom postojanja jugoslavenskih dr-

nja i porta, 14. travnja 2005. Njegova je temeljna zadaa

ava, premda je postojalo nekoliko pokuaja politikoga

sustavna i znanstveno utemeljena skrb o hrvatskome

nametanja zajednikoga imena jezika (srpsko-hrvatsko-

standardnome jeziku. Posebni zadatci Vijea su:

sloenaki za Kraljevine Jugoslavije; srpsko-hrvatski ili

hrvatsko-srpski za komunistike Jugoslavije). Za vrijeme

skrb o hrvatskome standardnome jeziku;

Drugoga svjetskoga rata i nekoliko godina nakon njega

raspravljati o aktualnim nedoumicama i otvorenim

svi su slubeni dokumenti u Jugoslaviji objavljivani na

pitanjima hrvatskoga standardnog jezika;

etiri slubena jezika (hrvatskome, makedonskome, slo-

upozoravati na primjere nepotivanja ustavne

venskome, srpskome), no ubrzo je mnogo po-litikoga

odredbe o hrvatskome kao slubenome jeziku u Re-

napora uporabljeno za ponovnu konvergenciju hrvat-

publici Hrvatskoj;

skoga i srpskoga. Unato svim pokuajima da se slubeno prizna postojanje hrvatskoga kao zasebnoga jezika,
nametanje zajednikoga nazivlja, vokabulara, pravopisa
i drugih jezinih normi u Jugoslaviji, dovelo je jedino
do slubenoga prihvaanja jednoga zajednikoga standardnoga jezika (srpsko-hrvatskoga) s dvije varijante (istonom ili srpskom i zapadnom ili hrvatskom). Reakcija
iz Hrvatske dola je ubrzo u obliku Deklaracije o nazivu
i poloaju hrvatskog knjevnog jezika koja se otvoreno

promicati kulturu hrvatskoga standardnog jezika u

pisanoj i govornoj komunikaciji;
skrbiti o statusu i ulozi hrvatskoga standardnoga jezika u svjetlu integracije Hrvatske u Europsku uniju;
donositi odluke u daljnjem procesu standardizacije
hrvatskoga standardnoga jezika;
brinuti o jezinim pitanjima i postavljati naela za
pravopisnu standardizaciju.

zalagala za priznavanje samostalnoga hrvatskoga jezika i

koju su jednoglasno 1967. potpisale vodee znanstvene,

Vijee za normu hrvatskoga standardnoga jezika sastaje

kulturne i obrazovne ustanove, kao i vodei intelektu-

se redovite i kroz temeljite rasprave dolazi do zakljuaka.

alci diljem Hrvatske, a koji su se tako otvorenim politi-

Institut za hrvatski jezik i jezikoslovlje udomljuje Vijee,

kim potezom nesumnjivo doveli u opasnost u komunis-

prua mu tehniku i administrativnu podrku kao i jezi-

tikim vremenima.

koslovne savjete kad je to potrebno. Institut za hrvatski jezik i jezikoslovlje [9] sredinja je hrvatska ustanova
za istraivanje hrvatskoga jezika, a jedan je od njegovih odjela (Odjel za hrvatski standardni jezik) posveen

Tijekom zadnjih 20 godina, etiri tokavski temeljena

opisu hrvatskoga standardnoga jezika s osobitom pozor-

standardna jezika razvijaju se samostalno kao nacionalni

nou na jezinu kulturu (npr. pruanje javnosti jezinih

standardni jezici u prirodno divergentnim smjerovima

savjeta ili pisanje jezinih prirunika). Savjeti o isprav-

budui da ne postoji nikakav sporazum ili koordinacija

noj jezinoj uporabi i jezikoslovna ekspertiza stalne su

glede njihovoga zajednikoga normiranja, pa su se time

dunosti Instituta. Savjeti se daju telefonski, e-potom

meu njima razlike uveale.

ili u pisanome obliku. Uz to, odgovori na najee pos-

19

tavljana pitanja dostupni su na portalu Jezini savjeti

manjina. Meutim, nije odreen kao obvezatan na sve-

[10] u sastavu institutova www-sjedita.

uilitima. Premda u Hrvatskoj postoje tenje, posebice

u prirodnim znanostima, da se predavanja odravaju na

Temeljna je zadaa Vijea za normu

hrvatskoga standardnoga jezika sustavna i
znanstveno utemeljena skrb o hrvatskome
standardnome jeziku.

engleskome jeziku, koje se tenje opravdavaju tvrdnjama

kako je to svrhovito i korisno, sasvim je jasno da bi bilo
iznimno tetno i neprihvatljivo ne pouavati na hrvatskome na sveuilitima. To bi imalo razarajui uinak
na razvoj hrvatskoga znanstvenoga nazivlja i strune fra-

Institutov projekt STRUNA [11], unutar kojega se ra-

zeologije. Stoga je Vijee za normu hrvatskoga stan-

zvija hrvatsko struno nazivlje zasluuje posebno spomi-

dardnoga jezika preporuilo Ministarstvu da slubeno

njanje. Cilj je ovoga projekta uspostava sustava koor-

odredi uporabu jezika u visokome obrazovanju.

dinacije terminolokih poslova u svim strunim podru-

U osnovnim i srednjim kolama Hrvatski jezik i knji-

jima u Hrvatskoj i time pripomoi poboljanju kakvoe

evnost pouava se kao predmet koji zauzimlje znaa-

i uinkovitosti viega obrazovanja i znanstvenih istra-

jan dio kolskih sati. Kao dio toga predmeta prouava se

ivanja izgradnjom jedinstvenoga provjerenoga nazivlja

hrvatska gramatika, rjenik i knjievnost, a razvijaju se

koje mogu rabiti strunjaci svih polja, a i zainteresirani

takoer pismeno i govorno izraavanje na hrvatskome

pojedinci i ope javnosti. Takoer se planira uspostava

jeziku. PISA testiranje, koje provjerava vjetine uenika

mree istraivakoga nazivlja kao i znanstvena suradnja

na svjetskoj razini, provodi se u Hrvatskoj od 2006, a

izmeu ustanova koje se bave razliitim vidovima termi-

prvi rezultati provjera pokazuju kako hrvatski petnaes-

nolokoga rada.

togodinjaci zauzimaju 26. mjesto na ljestvici svih zemalja svijeta i smjeteni su ispred uenika iz deset zemalja-

Danas su posuenice iz engleskoga jezika este u

govornome, a rjee u hrvatskome pisanom jeziku.

lanica EU i SAD.
U osnovnim i srednjim kolama uz hrvatski obvezatno je
uenje barem jednoga stranoga jezika od etvrtoga raz-

Pored toga, ostale hrvatske znanstvene ustanove (neko-

reda. Meutim, engleski se jezik (rijetko francuski ili

liko sveuilita s njihovim odsjecima za hrvatski jezik i

njemaki) nerijetko ue ve u djejem vrtiu. Engleski

knjievnost) i kulturne ustanove (kao to je Matica hr-

je uobiajeno prvi strani jezik u osnovnoj koli. Najra-

vatska) takoer sudjeluju u skrbi o hrvatskome jeziku.

ireniji drugi strani jezik je njemaki, potom slijede tali-

Javna glasila, kao to su dravna radio-televizija i neki

janski i francuski. U srednjim kolama ponekad se ue

novinski nakladnici imaju dobro razvijene korektorske i

ruski i panjolski kao drugi ili trei strani jezik. Latinski

lektorske slube za hrvatski standardni jezik, te obraaju

i starogrki ue se u klasiim programima koji poinju u

posebnu pozornost na kakvou jezika koji rabe u svojoj

petome razredu osnovne kole. K tome je latinski obve-

proizvodnji javno dostupnih tekstova.

zatan u svim humanistikim srednjim kolama. U koli

idovske manjine (koja ima pravo javnosti), mogue je

3.7 JEZIK U OBRAZOVANJU

uiti i hebrejski. Obrazovanje na manjinskim jezicima

dostupno je od djejega vrtia do srednje kole i hrvatska

Hrvatski je jezik sluben u svim osnovnim i srednjim

ga Vlada nancira za srpsku, eku, madarsku i talijan-

kolama osim u podrujima s puanstvom nacionalnih

sku manjinu.

20

3.8 MEUNARODNI ODNOSI

Najposjeenija hrvatska www-sjedita su: net.hr (portal za vijesti, sport, zabavu i zbivanja), index.hr (opi

Uporaba hrvatskoga standardnoga jezika u zemljama re-

www-portal, informacije, usluge, vijesti, sport, zabava,

gije regulirana je zakonima tih zemalja. Status hrvat-

vozila, gastronomija), jutarnji.hr (www sjedite dnevnih

skoga standardnoga jezika kao jednoga od slubenih je-

novinaJutarnji list), 24sata.hr (www-sjedite dnevnih

zika susjedne Bosne i Hercegovine od osobite je va-

novina 24 sata), tportal.hr (portal HT-a, Hrvatskih te-

nosti, pa hrvatske ustanove posveuju osobitu pozor-

lekomunikacija), njuskalo.hr (Njukalo portal s ogla-

nost suradnji sa znanstvenim i kulturnim ustanovama

sima), vecernji.hr (www-sjedite dnevnih novina Ve-

hrvatskoga naroda u Bosni i Hercegovini. Takoer, kul-

ernji list), forum.hr (najvei hrvatski www-forum na

turne ustanove iz Republike Hrvatske uspostavljaju su-

kojem se raspravlja o temama iz drutva, kulture, zabave

radnju s mnogim hrvatskim iseljenikim ustanovama di-

itd.). Sedam dnevnih novina svakodnevno objavljuje

ljem svijeta.

svoje lanke i na vlastitim www-sjeditima pored papirnatih izdanja.

Kad Republika Hrvatska pristupi Europskoj

uniji u 2013., hrvatski e jezik postati
24. slubeni jezik EU.

Rastua uloga Interneta vana

je i za jezine tehnologije.

Pouavanje hrvatskoga jezika organizirano je u inozem-

Institut za hrvatski jezik i jezikoslovlje odrava www-

nim kolama za djecu hrvatskih dravljana koji privre-

stranicu o hrvatskome jeziku koja donosi iscrpan popis

meno ili trajno ive u drugim zemljama. Hrvatski se

hrvatskih jedno- i viejezinih rjenika, gramatika i pra-

jezik pouava na mnogim inozemnim ustanovama i u

vopisnih prirunika. Na Filozofskome fakultetu Sve-

sreditima za slavenske jezike (tako postoji 36 slubenih

uilita u Zagrebu odrava se slina www stranica [13].

razmjenskih lektorata za hrvatski jezik i knjievnosti kao

Na istome se fakultetu od 1999. odrava i portal Jezine

i dva sredita za Hrvatske studije u Australiji i Kanadi

tehnologije za hrvatski jezik [14]. Wikipedija na hr-

koje sve podupire Ministarstvo znanosti, obrazovanja i

vatskome jeziku osnovana je 2003. i trenutano broji

porta Republike Hrvatske). Velik broj sredita za ue-

100.708 lanaka, te je 30. Wikipedija po slubenome

nje hrvatskoga kao drugoga ili stranoga jezika postoji u

broju lanaka.

Hrvatskoj, a najpoznatiji je Croaticum [12].

Pristup jezinim resursima za hrvatski jezik u zadnje je

vrijeme olakan zbog broja hrvatskih ustanova i organi-

3.9 HRVATSKI NA INTERNETU

zacije koje provode postupke digitalizacije (ukljuujui

znaajne projekte koje podupire Ministarstvo znanosti,

Statistika Dravnoga zavoda za statistku o uporabi in-

obrazovanja i porta i Ministarstvo kulture u digitaliza-

formacijskih i komunikacijskih tehnologija u poduze-

ciji hrvatske kulturne batine), a koja je uveala vidljivost

ima i kuanstvima dana je na slikama 6 i 7.

hrvatskoga jezika meu ostalim internetskim izvorima.

21

Uporaba informacijskih i komunikacijskih tehnologije (ICT) u poduzeima (%)

uporaba raunala
pristup Internetu
www-sjedite
uporaba nancijskih i bankovnih usluga
uporaba usluga e-uprave

2008

2009

2010

98
97
64
84
56

98
95
57
84
61

97
95
61
85
63

6: ICT u poduzeima

Kuanstva opremljena informacijskim i komunikacijskim tehnologijama (ICT) (%)

osobno raunalo
pristup Internetu
mobilni telefon

2008

2009

2010

53
45
81

55
50
82

60
57

7: ICT u kuanstvima

22

4
JEZINOTEHNOLOKA
PODRKA ZA HRVATSKI
Jezine se tehnologije koriste za razvoj sustava namije-

raunalno potpomognuto uenje jezika

njenih za obradbu prirodnoga jezika, te se mogu pojaviti

pretraga obavijesti

i pod nazivom prirodnojezine tehnologije. Prirodni

se jezik pojavljuje u govorenom i pisanom obliku. Dok

crpljenje obavijesti

je govor najstariji i najprirodniji oblik jezinoga priop-

saimanje teksta

avanja, sloene obavijesti i veina ljudskoga znanja sa-

odgovaranje na pitanja

drana je i prenosi se s pomou teksta. Govorne tehno-

prepoznavanje govora

logije i tehnologije obradbe teksta obrauju jezik u ovim

dvama nainima njegova ostvaraja sluei se rjenicima,
gramatikim pravilima i znaenjem. To znai da jezine
tehnologije povezuju jezik s razliitim oblicima znanja,
neovisno o mediju (govoreni ili pisani tekst) u kojem se
ostvaruje. Slika 8 prikazuje okruje jezinih tehnologija.
U naem priopavanju mijeamo jezik s ostalim oblicima priopavanja i drugim medijima, npr. govor ukljuuje gestikulaciju i mimiku. Digitalni tekst povezujemo

generiranje govora
Jezine su tehnologije ve vrsto uspostavljeno zasebno
istraivako podruje s irokim rasponom uvodne literature. Zainteresirani itatelj se upuuje na sljedee referencije: [15, 16, 17, 18, 19].
Prije nego to krenemo prikazivati navedena podruja istraivanja, na kratko bismo opisali arhitekturu uobiajenoga jezinotehnolokoga sustava.

sa slikama i zvukovima. U lmovima se jezik pojavljuje

u govorenom i pisanom obliku. Stoga se govorne tehnologije i tehnologije obradbe teksta preklapaju i proimaju se s mnogim drugim tehnologijama koje pospjeuju obradbu multimedijskoga priopavanja i multimedijskih tehnologija.

4.1 ARHITEKTURE
JEZINOTEHNOLOKIH
APLIKACIJA

U ovome emo poglavlju prikazati glavna podruja pri-

Uobiajena se programska aplikacija za obradbu jezika

mjene jezinih tehnologija, kao to su jezina provjera,

sastoji od nekoliko dijelova koji se bave razliitim jezi-

www-trailice, govorna interakcija i strojno prevoenje.

nim slojevima. Dok takve aplikacije mogu biti vrlo slo-

Te aplikacije i temeljnje tehnologije ukljuuju:

ene, slika 9 pokazuje znatno pojednostavnjenu arhitekturu kakva se moe nai u sustavima za obradbu teksta.

provjeru pravopisa

Prva tri modula bave se strukturom i znaenjem ulaz-

potpora stvaranju tekstova

noga teksta:

23

Govorne tehnologije
Multimedijske i
multimodalne
tehnologije

Jezine
tehnologije

Tehnologije znanja

Tekstne tehnologije

8: Jezine tehnologije

1. predobradba: ienje ulaznih podataka, analiza i

i mnoge druge. Nakon uvodnoga dijela o osnovim po-

uklanjanje oblikovanja teksta, odreivanje ulaznoga

drujima primjene jezinih tehnologija, dat e se kra-

jezika, ponekad umetanje nedostajuih dijakritikih

tak pregled stanja jezinih tehnologija u istraivanjima

znakova u hrvatskome, itd.

i obrazovanju koji e se zakljuiti pregledom prolih, sa-

2. gramatika ralamba: pronalaenje glagola i njego-

danjih i buduih istraivakih programa razvoja jezi-

vih objekata, subjekata, njihovih atributa itd.; pre-

nih tehnologija za hrvatski jezik [20]. Na kraju ovoga

poznavanje reenine strukture.

poglavlja prikazat e se struna procjena stanja osnov-

3. semantika ralamba: razoblienje (npr. koje znae-

nih jezinih resursa i alata za hrvatski jezik sagledanoga

nje rijei glava je primjereno u danom kontekstu?),

kroz niz kategorija kao to su dostupnost, zrelost ili kak-

razrjeenje anafore (odreivanje na to se tono od-

voa. Ope stanje jezinih tehnologija za hrvatski jezik

nosne zamjenice kao to su ona, kojemu, itd. u

je saeto u obliku tablice. Najvaniji resursi i alati koji

tekstu odnose); predstavljanje znaenje reenice u

se opisuju u tekstu su dani masnim slovima, a moe ih se

strojno itljivome obliku.

takoer nai na slici 15 na kraju poglavlja. Jezine tehnologije za hrvatski jezik usporeene su i s jezinim tehno-

Nakon te analize zadatkovno-orijentirani moduli izvode

logijama za druge jezicima ukljuene u niz ovih bijelih

mnoge specine postupke kao to su npr. automat-

knjiga.

sko saimanje ulaznoga teksta, pretraga baza podataka

Ulazni tekst

Predobradba

Izlaz

Gramatika analiza

Semantika analiza

Moduli za posebne
zadatke

9: Tipina aplikacija za obradu teksta

24

4.2 OSNOVNA PODRUJA

PRIMJENE JEZINIH
TEHNOLOGIJA
4.2.1 Jezina provjera

mnogo ei niz dviju rijei nego jaz generacija. Statistiki se jezini model moe proizvesti automatski iz velike koliine (ispravnih) jezinih podataka (tj. iz korpusa). Do sada su se ova dva pristupa mahom razvila i
provjerila na jezinim podatcima za engleski jezik. Meutim, na hrvatski jezik takva rjeenja nisu izravno pri-

Svatko tko koristi obradnike teksta kao to je npr. Mi-

mjenljiva zbog njegove bogate eksije i slobodnijega

croso Word, naiao je pravopisni provjernik koji obi-

reda rijei u reenici koji u mnogome pridonose takvim

ljeava pogrjeke u tipkanju i predlae ispravke. Prvi pra-

sustavima problematinoj rasprenosti podataka.

vopisni provjernici usporeivali su rijei iz teksta s rje-

Uporaba jezinih provjernika nije ograniena samo na

nikom ispravno napisanih rijei. Danas su ovi programi

obradnike teksta, ve se koristi i u potpornim alatima

znatno razraeniji. Uz dodatak jezino ovisnih algori-

za stvaranje teksta kao to su opseni prirunici i ostala

tama za obradbu morfologije (npr. prepoznavanje raz-

tehnika dokumentacija kad je rije o primjeni raunal-

liitih padea), neki ve mogu prepoznati i sintaktike

nih sustava u informacijskim tehnologijama, zdravstvu,

pogrjeke kao to je isputanje glagola ili nesronost iz-

strojarstvu i drugdje. Bojei se korisnikih prituaba o

meu subjekta i predikata u broju i rodu, npr. Ona

pogrjenoj uporabi ili odtetnih zahtjeva zbog nepreciz-

je *pisao pismo. Pa ipak, i najnapredniji provjernici ne

nih ili loe shvaenih korisnikih uputa, tvrtke se sve

mogu pronai pogrjeke u prvoj stro pjesme Jerrolda

vie okreu prema stvaranju to kvalitetnijih korisni-

H. Zara (1992):

kih uputa i tehnike dokumentacije dok se istodobno

pokuavaju iriti na meunarodnome tritu (kroz pre-

I have a spelling checker,

voenje i lokalizaciju). Napredak u raunalnoj obradbi

It came with my PC.

prirodnoga jezika doveo je do razvoja potpornih pro-

It plane lee marks four my revue

grama za pisanje teksta koji pomau piscima tehnike

Miss steaks aye can knot sea.

dokumentacije pri uporabi kontroliranoga jezika u ko-

Za razrjeivanje ovakvih pogrjeaka u mnogim je sluajevima potrebna ralamba konteksta, npr. treba li u hrvatskome imenicu pisati velikim (ensko osobno ime) ili
malim (opa imenica) poetnim slovom kao u sluaju:
Slatka je ova vinja. [is cherry is sweet.]
Slatka je ova Vinja. [is Cherry is sweet.]

jem je u skladu s (korporativnim) pravilima ograniena

uporaba leksikih jedinica, strunoga nazivlja ili jednostavijih sintaktikih struktura. Za hrvatski takvi alati jo
nisu na raspolaganju.

Jezini provjernici nisu ogranieni samo

na obradnike teksta, ve se koriste i u potpornim
alatima za stvaranje teksta.

Takva ralamba zahtijeva bilo oblikovanje jezino posebnih gramatikih pravila, izradba kojih ukljuuje vi-

Premda su istraivanja raunalnih modela hrvatske ek-

soku strunost i mnogo radnih sati, ili uporabu tzv. sta-

tivne morfologije postojala jo u 1980-ima, prvi je ko-

tistikih jezinih modela. Takvim je modelima tono

mercijalni prvopisni provjernik Hrvatski raunalni pra-

izraunana vjerojatnost pojavljivanja neke rijei u odre-

opis objavljen tek 1996. [8] Ubrzo ga je preuzeo Mi-

enome kontekstu (tj. s obzirom na prethodeu ili sli-

croso i danas je sastavni dio Microso Ocea te je u

jedeu rije). Na primjer, u hrvatskome je jaz izmeu

najiroj uporabi. Nekoliko je privatnih tvrtki takoer

25

Statistiki jezini modeli

Ulazni tekst

Pravopisna provjera

Gramatika provjera

Prijedlozi ispravaka

10: Tipina aplikacija za jezinu provjeru

razvilo pravopisne provjernike, ali niti jedan nije bio to-

hvaenih rezultata nisu se znaajno promijenili od prve

liko uspjean. On-line Hrvatski akademski spelling chec-

inaice. U trenutanoj inaici Google nudi pravopisnu

ker (Hascheck) [21] postoji od 1994 i jo uvijek je u

provjeru za pogrjeno napisane rijei, a ukljuuje u pre-

uporabi. Za hrvatski takoer postoji i besplatni prvo-

tragu i osnovne semantike elemente kojima je mogue

pisni provjernik temeljem na ispell/aspell aplikaciji, a

poboljati tonost pretrage analizom znaenja upita u

uporabiv je na svim platformama na kojima je dostu-

danome kontekstu [23]. Uz pomo ovoga algoritma

pan OpenOce. Svi se ovi programi temelje na vrlo

Google je poeo pokrivati hrvatske rijei u nekima od

velikim leksikonima ispravno napisanih rijei, a taj pris-

oblika u kojima se pojavljuju u tekstu. Za razliku od

tup ima dvije osnovne manjkavosti: 1) nizovi pismena

npr. engleskih imenica gdje postoje samo etiri mogua

koji predstavljaju ispravno napisane rijei mogu se po-

oblika (hand, hands, hands, hands), hrvatske se te-

javiti u pogrjenome kontekstu; 2) nemogunost pre-

oretski mogu pojaviti u 14 razliitih oblika, ali su pro-

poznavanja ispravno napisanih rijei koje su nepoznate

sjeno predstavljeni s 10 razliitih nizova (ruka, ruke,

leksikonu. Osim provjernika pravopisa i potpornih pro-

ruci, ruku, rukom, rukama ...). Googleova trailica moe

grama za stvaranje teksta, jezina je provjera vana i na

pronai oblike kao to su ruka ili ruke, ali oblik ruci ve

podruju strojno potpomognutoga uenja jezika, a pri-

nije vie povezan uz imenicu ruka. Ima jo dosta pros-

mjenjuje se i kod automatskoga ispravljanja upita posla-

tora za poboljanja u Googleovoj trailici kod ektivno

nih www-trailicama, npr. Googleove Jeste li mislili...

bogatih jezika kod kojih se mora nositi s injenicom da

preporuke.

se pojedine rijei mogu pojaviti u veem broju oblika.

Meutim, uspjeh Googlea pokazuje kako s golemim ko-

4.2.2 WWW trailice

Pretraga na mrei, na intranetima ili u digitalnim knjinicama danas je vjerojatno najire koritena, a ipak jo
nedovoljno razvijena, jezina tehnologija. Trailica Google, koja je zapoela 1998., danas se rabi za otprilike

liinama podataka i uinkovitim tehnikama njihova indeksiranja, preteito statistiki utemeljeni pristup moe
dovesti do zadovoljavajuih rezultata, no njihova kakvoa takoer ovisi i o samoj strukturi prirodnoga jezika
na kojem se pretrauje.

80% svih pretraga u svijetu [22]. Od 2004. u hrvat-

Pa ipak, za razraenije pretrage obavijesti ukljuivanje

skome se koristi glagol guglati/googlati i njegove tvore-

dubljega jezinoga znanja bit e kljuno za ispravnu in-

nice (iz-/na-/pre-/pro-/u-)guglati/(iz-/na-/pre-/pro-/u-

terpretaciju rezultata. Eksperimenti u kojim se rabe lek-

)googlati premda si jo nije izborio mjesto u tiskanim

siki resursi kao to su strojno itljivi tezaurusi i onto-

rjenicima (zabiljeeni su ak i sloenije tvorenice kao

loki organizirani jezini resursi (npr. WordNet za en-

npr. ugugljiv). Ni suelje za pretragu, niti prikaz do-

gleski ili Hrvatski Wordnet CroWN) pokazuju ozbi-

26

Mrene stranice

Predobradba

Semantika obradba

Indeksiranje
Pronalaenje
i
relevantnost

Predobradba

Analiza upita

Korisniki upit

Rezultati pretrage

11: Arhitektura www-trailice

ljan napredak omoguujui pronalaenje www-stranica

rijei. Za upit Daj mi popis svih tvrtki koje su bile pre-

na temelju sinonima upitnih rijei, npr. nuklearna ener-

uzete od drugih tvrtki u zadnjih pet godina, potrebna je

gija i atomska energija ili na temelju rijei jo udaljenije

sintaktika i semantika analiza. Sustav takoer mora

povezanih s upitnim rijeima.

urno priskrbiti i tako indeksirane dokumente. Za zadovoljavajui odgovor na ovo pitanje potrebna je pri-

Sljedei e narataj trailica morati ukljuivati

razraenije jezine tehnologije.

mjena sintaktike ralambe (parsanja) kako bi se ralanila sintaktika struktura reenice i odredilo kako korisnik trai tvrtke koje je neka tvrtka preuzela, a ne tvrtke

Pa ipak, za razraenije pretrage obavijesti ukljuivanje

dubljega jezinoga znanja bit e kljuno za ispravnu interpretaciju rezultata. Eksperimenti u kojim se rabe leksiki resursi kao to su strojno itljivi tezaurusi i ontoloki organizirani jezini resursi (npr. WordNet za engleski ili Hrvatski Wordnet CroWN) pokazuju ozbiljan napredak omoguujui pronalaenje www-stranica
na temelju sinonima upitnih rijei, npr. nuklearna energija i atomska energija ili na temelju rijei jo udaljenije
povezanih s upitnim rijeima.
Sljedei e narataj trailica morati ukljuivati razraenije jezine tehnologije, osobito ako upit bude sadravao
pitanje ili kakvu drugu reenicu umjesto popisa upitnih

koje su preuzele neku tvrtku. Takoer, izraz u zadnjih

pet godina, treba obraditi kako bi se odredilo o kojih
je tono pet godina rije uzimajui u obzir tekuu godinu. Konano, obraeni upit se mora sraziti s golemom
koliinom nestrukturiranih podataka kako bi se pronaao komadi ili komadii obavijesti koje korisnik trai.
Taj se zadatak obino naziva pretraga obavijesti i ukljuuje pretragu i rangiranje relevantnih dokumenata. K
tome pri sastavljanju zahtijevanoga popisa tvrtki, sustav
mora moi prepoznati u pretraivanim dokumentima
kako odreen niz pismena doista predstavlja ime tvrtke.
Takav se zadatak zove prepoznavanje imena i obavlja ga
specijalizirana aplikacija za tu namjenu. Jo su zahtjev-

27

niji pokuaji da se na temelju upita pronau relevantni

dokumenti pisani na drugim jezicima. Za takvo viejezino pretraivanje obavijesti moramo strojno prevesti
upit na se mogue jezike i pronaene dokumente prevesti na jezik upita.
Rastua koliina podataka dostupnih u netekstnim
oblicima potie stvaranje usluga koje bi omoguile multimedijsko pretraivanje obavijesti, npr. pretragu u slikovnim, audio- ili video-zapisima. Za audio- i videozapise jo je potreban i modul za raspoznavanje govora
koji bi omoguio pretvorbu govora u tekst ili njegov fonetski zapis u kojem se onda moe obavljati pretraga.
Za ektivno bogate jezike kao to je hrvatski, trailice
moraju omoguiti pretragu odjednom po svim oblicima
u kojima se neka rije moe pojaviti, umjesto da se svaki
oblik mora unositi pojedinano. Takav oblik pretrage
mogue je izvesti s pomou Hrvatskoga lematizacijskoga
posluitelja koji je razvijen na Odsjeku za lingvistiku Filozofskoga fakulteta Sveuilita u Zagrebu i slobodno

4.2.3 Govorna interakcija

Govorna interakcija jedno je od mnogih podruja primjene govornih tehnologija, tj. tehnologija za obradbu
govora. Tehnologije za govornu interakciju stvaraju suelja koja omoguuju komunikaciju govorenoga jezika
umjesto grakoga suelja, tipkovnice ili mia. Danas
su takva govorna korisnika suelja (oice user interfaces, VUI) ukljuena u djelomino ili potpuno automatizirane usluge koje razne tvrtke nude svojim korisnicima, zaposlenicima ili partnerima putem telefona. Podruja koja danas u mnogome ovise o uporabi VUI-ja su
bankarstvo, logistika, javni prijevoz i telekomunikacije.
Drugi oblici uporabe tehnologija za govornu interakciju su suelja prema pojedinim ureajima, npr. sustavi
za cestovnu navigaciju ili sustavi gdje je govorna interakcija zamjena za ulazno/izlazne podatke grakih suelja,
npr. kod pametnih telefona.
Sustavi koji koriste tehnologiju za govornu interakciju
sastoje se od etiri razliita podsustava s pripadajuim
tehnologijama:

je dostupan preko Interneta [24] omoguujui pristup

1. Automatsko prepoznavanje govora (automatic spe-

Hrvatskome morfolokome leksikonu, opsenoj bazi

ech recognition, ASR) odreuje koje su rijei zaista iz-

podataka hrvatskih rijei i svih njihovih oblika. Ta baza

govorene u nizu glasova koje je korisnik izrekao.

sadri preko 110.000 natuknica iz kojih je generirano

2. Razumijevanje prirodnoga jezika bavi se analizom

preko 4 milijuna oblika tako da svaki zapis u bazi sadri

sintaktike strukture korisnikova iskaza i njegovom

natuknicu, oblik i MSD-oznaku tj. popis svih grama-

interpretacijom u skladu s namjenom odreenoga

tikih kategorija koje su se ostvarile tim oblikom. Taj je

sustava.

zapis usklaen s MulText East [25] preporukama.

3. Upravljanje razgovorom odreuje koju akciju sustav

mora poduzeti na temelju korisnikove upute i na te-

Godine 2009., kao rezultat zajenikoga amanskohrvatskoga projekta CADIAL [26], vladina agencija
HIDRA omoguila je javni www-pristup svim hrvatskim zakonskim i podzakonskim dokumentima putem
ektivno osjetljive trailice [27]. Ta trailica takoe

melju ukupne funkcionalnosti toga sustava.

4. Sinteza govora (text-to-speech, TTS) tehnologija se
rabi za pretvaranje pojedinih rijei nekoga iskaza u
niz glasova koji e biti odaslani korisniku.

omoguuje i viejezinu pretragu dokumenata s obzi-

Jedan od glavnih izazova jest kako da ASR-sustav to

rom na to da su svi dokumenti indeksirani deskripto-

tonije prepozna rijei koje je korisnik uporabio. To

rima iz EUROVOC-a to omoguuje uporabu i engle-

zahtijeva ili ogranienje broja moguih korisnikih is-

skih deskriptora u upitu.

kaza na ogranien skup rijei, ili runo stvaranje jezi-

28

Govorni izlaz

Govorni ulaz

Govorna sinteza

Obradba signala

Fonetski odabir i
planiranje intonacije

Razumijevanje
prirodnoga jezika i
dijalog

Prepoznavanje

12: Govorna interakcija

noga modela koji pokriva irok raspon moguih iskaza

dananje sustave za TTS mogue je podeavati do e-

na prirodnome jeziku. Uporabom postupaka strojnoga

ljene kakvoe s obzirom na prirodnost naglaska dina-

uenja jezini se modeli mogu automatski generirati iz

mino organiziranih iskaza.

govornih korpusa, tj. velikih zbirki govornih audio sni-

U proteklom je desetljeu na tritu tehnologija za go-

maka i njihove tekstovne transkripcije. Ograniavanje

vornu interakciju dolo do uznapredovale standardiza-

skupa doputenih iskaza obino ne omoguuje korisni-

cije sueljavanja razliitih tehnolokih komponenta. Ta-

cima uporabu VUI-ja na prirodan nain, pa takvi sus-

koer je u proteklome desetljeu dolo do znaajnoga

tavi znaju korisnicima izgledati i zvuati odbojno. Is-

okrupnjivanja na tritu, osobito kad je rije o ASR i

todobno, sastavljanje, podeavanje i odravanje takvih

TTS sustavima. Na tritima u zemljama skupine G20

opsenih jezinih i govornih modela znatno poskup-

tj. gospodarski jakih zemalja znaajne populacije, prevla-

ljuje takve sustave. S druge strane sustavi koji rabe je-

dava pet svjetski relevantnih tvtrki, s tim to su Nuance

zine modele, ve u polasku doputaju korisnika slobod-

(SAD) i Loquendo (Italija) najzastupljenije u Europi.

nije izraavanje, npr. zapoinjanjem razgovora reeni-

Godine 2011. Nuance je najavio preuzimanje Loquenda

com Kako vam mogu pomoi?, pokazuju i vii stupanj

to e znaiti daljnji korak u okrupnjavanju trita.

automatizacije i vii stupanj korisnikoga prihvaanja.

Premda je baza hrvatskih difona razvijena jo 1998. unutar projekta MBROLA [28] u kojem je sudjelovao Od-

Govorna interakcija je temelj za stvaranje suelja

koja omoguuju korisniku uporabu govora
umjesto grakoga suelja, tipkovnice i mia.

sjek za fonetiku Filozofskoga fakulteta Sveuilita u Zagrebu, do danas jo uvijek ne postoji niti jedan komercijalni sustav za hrvatski ATS ili TTS razvijen u Hrvatskoj.
Istraivanja na ovome podruju provode se i na Fakul-

Za odreene dijelove VUI-ja, tvrtke nerijetko rabe

tetu elektrotehnike i raunarstva istoga sveuilita [29],

snimljene iskaze profesionalnih spikera. Tako snimljeni

ali i na Sveuilitu u Rijeci jaka skupina istraivaa radi

statini iskazi, u kojima se rijei ne mijenjaju ovisno o

na razvoju resursa i alata za obradbu hrvatskoga govora

kontekstu uporabe ili o osobnim podatcima pojedinoga

[30, 31, 32].

korisnika, korisniku mogu pruiti kvalitetu govora koju

Ako bi se pokuao baciti pogled onkraj sadanjega sta-

oekuje. Meutim, to je sadraj dinaminiji, tada je i

nja ove tehnologije, mogle bi se oekivati znaajne pro-

vjernost govora manja jer je potrebno umjetno povezi-

mjene s obzirom na ubrzano irenje pametnih telefona

vati velik broj malih snimki. Nasuprot ovim sustavima,

kao nove platforme za odnose s korisnicima uz ve pos-

29

tojee kao to su telefon, Internet i e-pota. Ovakav se

marku) ili na razini smjetanja prijedlonoga izraza u

razvoj prilika moe oekivati i u sluaju primjene teh-

sintaktikoj strukturi kao u primjeru:

nologije za govornu interakciju. S jedne e strane i dugorono gledano potrebe za klasinim telefonskim VUI
zacijelo opadati, a s druge e strane uporaba govora kao
izvora ulaznih podataka za pametne telefone zacijelo
biti u porastu. Taj smjer razvoja takoer se moe prepoz-

Policajac je uoio ovjeka bez teleskopa.

[e policeman spotted a man without a telescope.]
Policajac je uoio ovjeka bez pitolja.
[e policeman spotted a man without a pistol.]

nati s obzirom na vidan napredak tonosti prepoznava-

Jedan od moguih pristupa izgradnji strojnoprevoditelj-

nja govora neovisnoga o govorniku u sustavima za dikti-

skih sustava temeljen je na jezinim pravilima. Za pre-

ranje koji se ve nude kao usluga korisnicima pametnih

voenje izmeu bliskosrodnih jezika, izravno bi prevo-

telefona.

enje moglo biti lake izvedivo. Nerijetko sustavi temeljeni na pravilima (ili na jezinome znanju) ralanjuju

4.2.4 Strojno prevoenje

ulazni tekst i pretvaraju ga u posrednu simboliku pre-

Zamisao uporabe raunala za prevoenje s jednoga pri-

zentaciju iz koje se onda generira tekst na ciljnome je-

rodnoga jezika na drugi moe se smjestiti jo u 1946.,

ziku. Uspjeh ovakvih pristupa u mnogome ovisi o dos-

a uslijedila joj je znaajna potpora za istraivanja u tom

tupnim opsenim rjenicima s morfolokim, sintakti-

podruju tijekom 1950-ih i ponovno u 1980-ima. Pa

kim i semantikim podatcima, ali i o velikom broju gra-

ipak, strojno prevoenje (machine translation, MT) jo

matikih pravila koje su briljivo izradili visokostruni

uvijek ne uspijeva ispuniti visoka oekivanja glede nje-

jezikoslovci. To je vrlo zahtjevan, dugotrajan i stoga

gove kakvoe.

skup posao.
S krajem 1980-ih, kad je porasla snaga raunala i kad su
ona postala dostupnija, vie se zanimanja poelo posve-

U svom najjednostavnijem obliku MT

samo zamjenjuje rijei jednoga prirodnoga
jezika rijeima iz drugoga.

ivati statistikim modelima u strojnome prevoenju.

Statistiki modeli izvedeni su iz ralambe dvojezinih
korpusa tj. usporednih korpusa kao to je npr. Europarl
korpus, koji sadri zapisnike sjednica Europskoga parla-

Najjednostavniji oblik strojnoga prevoenja sastoji se od

menta na 21 europskih jezika, ili JRC-Acquis usporedni

zamjene rijei jednoga prirodnoga jezika rijeima iz dru-

korpus [33] na 22 europska jezika. Kad im se osigura do-

goga. To moe biti uporabivo u uskim podrujima s izra-

voljno podataka, statistiki sustavi za strojno prevoenje

zito ogranienim, formulainim izrazima, npr. kod vre-

rade dovoljno dobro za dobivanje priblinoga znaenja

menskih izvjea. Meutim, za dobar prijevod manje

teksta na stranome jeziku obradbom usporednih teks-

ogranienih tekstova, vei tekstni odsjeci (fraze, ree-

tova i pronalaenjem odgovarajuih prijevodnih podu-

nice ili itavi odlomci) moraju se to vie u prijevodu pri-

darnosti meu njima. Meutim, za razliku od sustava

bliiti svojim prijevodnim ekvivalentima u polaznome

temeljenih na znanju (ili pravilima), MT sustavi teme-

jeziku. Najvei je problem u tome to je prirodni je-

ljeni na statistici (ili podatcima) esto generiraju tekst

zik vieznaan, a taj se problem pojavljuje na vie razina,

koji nije ovjeren tj. nije usklaen s gramatikom ciljnoga

npr. na razini razoblienja znaenja rijei (word sense di-

jezika. Podatkovno temeljeni pristupi strojnome prevo-

sambiguation, WSD) na leksikoj razini ( Jaguar na po-

enju su u prednosti jer zahtijevaju manje ljudskoga na-

etku reenice moe znaiti ivotinju ili automobilsku

pora, a mogu pokriti i osobitosti jezika (npr. idiomatske

30

Izvorni tekst

Analiza teksta (oblikovanje,

morfologija, sintaksa, itd.)

Statistiko
strojno
prevoenje

Prijevodna pravila
Ciljni tekst

Generiranje teksta

13: Statistiko strojno prevoenje

izraze) koje obino sustavi temeljeni na pravilima zane-

esto postavlja i zahtjev za sronou glede tih kategorija

maruju ili zaobilaze. Kad se gledaju samo europski je-

izmeu npr. atributa i imenice u rodu, broju i padeu ili

zici, prihvatljivi se prijevodi mogu dobiti za engleski i

samo u rodu i broju kad je rije o subjektu i predikatu.

romanske jezike, no kakvoa prijevoda znaajno opada

za ostale germanske, slavenske, ugronske ili baltike jezike [34].
Kako su prednosti i nedostatci sustava za strojno prevo-

Strojno je prevoenje osobito izazovno za

slavenske jezike zbog njihova slobodnoga reda
rijei, bogatstva oblika rijei i postojanja
udaljenih a meuovisnih dijelova iste fraze.

enje temeljenih na znanju i sustava temeljenih na podatcima upravo komplementarno rasporeeni, danas se

Premda su eljko Bujas i Bulcs Lszl jo 1959. orga-

istraivai mahom usmjeravaju na hibridne pristupe koji

nizirali prvu radionicu o strojnome prevoenju [35] na

kombiniraju metodologije obiju vrsta ovih sustava. Je-

Filozofskome fakultetu Sveuilita u Zagrebu, nikakvo

dan od naina hibridizacije jest uporaba obje vrste sus-

ozbiljnije istraivanje o strojnome prevoenju za hrvat-

tava za obavljanje prevoenja, a potom selekcijski mo-

ski jezik nije se dogodilo prije 21. stoljea. Projekt In-

dul odluuje o tome koji je rezultat vie kakvoe za svaku

formacijske tehnologije u prevoenju i e-uenju hrvat-

pojedinu reenicu. Na alost, u sluaju duljih reenica,

skoga [36] pokrenut je 2007. s ciljem istraivanja koji su

npr. vie od 12 rijei, niti jedan sustav jo uvijek ne daje

preduvjeti potrebni za stvaranje MT sustava za prevoe-

prijevod eljene kakvoe. Uinkovitijim se doima pris-

nje na i s hrvatskoga jezika. Poevi od 2010. Europska

tup u kojem se kombiniraju najbolji dijelovi reenica iz

je komisija pokrenula i potpomae nekoliko projekata

viestrukih moguih prijevoda, a oni mogu biti popri-

kako bi se razvila istraivanja i razvoj strojnoga prevo-

lino sloeni s obzirom da odgovarajui dijelovi vies-

enja za tzv. jezike s nedovoljno razvijenim resursima, a

trukih prijevodnih rjeenja nisu uvijek lako uoljivi, te

meu njih je ukljuen i hrvatski. Tako CIP ICT PSP

se moraju posebno sravnjivati.

projekt LetsMT! [37] i FP7 projekt ACCURAT [38]

Strojno je prevoenje s i na hrvatski jezik osobito iz-

razvijaju nove metode za to jednostavnije prikupljanje

azovan zadatak. Slobodniji red rijei u reenici i bo-

podataka potrebnih za strojno prevoenje i izgradnju

gata eksija predstavljaju probleme pri generiranju is-

takvih sustava prilagoenih razliitim domenama i obli-

pravnih reeninih konstrukcija i oblika rijei koji obli-

cima primjene. U oba ova projekta kao hrvatski part-

nim nastavcima kodiraju gramatike kategorije roda, pa-

ner sudjeluje skupina istraivaa s Filozofskoga fakulteta

dea, broja, naina, vremena itd. Dodatne probleme

Sveuilita u Zagrebu.

31

Projekt ACCURAT [39] istrauje nove metode upo-

voenje u kojima se ve rabe velike terminoloke baze

rabe usporedivih korpusa ne bi li se nadoknadila nesta-

i prijevodne memorije. Dodatni je problem to je veina

ica jezinih resursa i posredno poboljalo strojno prevo-

trenutanih strojnoprevoditeljskih sustava usmjerena na

enje za jezike s nedovoljnim resursima i za uske domene

engleski i podupire svega jo nekoliko jezika pri prevoe-

[40]. Cilj je projekta ACCURAT postii znaajan na-

nju na hrvatski i s hrvatskoga. To zapravo onemoguuje

predak u kakvoi strojnoga prijevoda za itav niz novih

druge prijevodne smjerove, a istodobno zahtijeva od ko-

slubenih jezka EU i jezika zemalja-pristupnica (eston-

risnika da se slue veim brojem raznorodnih sustava.

ski, grki, hrvatski, letonski, litavski i rumunjski), kao

Postupci vrjednovanja omoguuju usporeivanje kak-

i predloiti nove pristupe za prilagodbu tehnologija za

voe prijevoda sustava za strojno prevoenje, njihove

strojno prevoenje u pojedinim uskim domenama i time

razliite pristupe i stanje strojnoprevoditeljskih sustava

znaajno poveati pokrivenost razliitih jezika i domena

za razliite jezike. U okviru projekta Euromatrix+ sas-

strojnim prevoenjem.

tavljena je 14 u kojoj je prikazana kakvoa za sve parove

Projekt LetsMT! [41] izgrauje novi vrstu on-line su-

strojnih prijevoda izmeu 22 slubena jezika EU-a (irski

radne platforme za dijeljenje usporednih tekstova i auto-

jedino nedostaje) iskazana s pomou BLEU mjere [33]

matsko stvaranje vlastitih sustava za strojno prevoenje.

koja s veim brojem bodova iskazuje viu kakvou pri-

Ova platforma smjetena u raunalnome oblaku omo-

jevoda. Ljudski prevoditelji obino postiu oko 80 bo-

guit e svim vrstama korisnika slanje u posebno zati-

dova.

en repozitorij vlastitih jezinih resursa na temelju ko-

Najbolji rezultati (prikazani zeleno i plavo) postignuti

jih e se potom automatski izraditi vlastiti sustav za sta-

su za jezike za koje ve postoje sustavi znatno razraeni

tistiko strojno prevoenje treniran upravo na temelju

unutar raznih istraivakih programa i za koje jezike

tih vlastitih jezinih resursa. Takav sustav za strojno

postoji mnogo usporednih korpusa (npr. engleski, fran-

prevoenje potom se moe podijeliti s ostalim korisni-

cuski, nizozemski, panjolski, njemaki), a najgori su re-

cima. Strojnoprevoditeljske usluge projekta LetsMT!

zultati (u crvenome) za jezike koji su u mnogome struk-

mogu se rabiti na nekoliko naina: kroz www-portal,

turno razliiti od veine jezika (npr. madarski, maltski,

kroz widget koji se moe slobodno preuzeti i ukljuiti na

nski).

svoje www-stranice, kroz dodatak za popularne prebirnike kao i kroz integraciju u postojee sustave za strojno
potpomognuto prevoenje kako on-line, tako i o-line.

4.2.5 Ostala podruja primjene

Izgradnja jezinotehnolokih aplikacija ukljuuje itav

Google Translate nudi prijevode na hrvatski i s hrvat-

niz podzadataka koji se ne vide na razini korisnikoga

skoga od 2008. Kakvoa njegovih prijevoda bila je niska

suelja, ali ispod poklopca osiguravaju funkcionalnost

u poetku, ali se poboljava kako je sve vie i vie uspo-

toga sustava. Upravo ti podzadatci rezultat su rjeavanja

rednih hrvatsko-engleskih podataka dostupno on-line.

vanih istraivakih problema koji su prerasli u pojedine

Jo uvijek se smatra kako upravo na podruju pobolja-

podgrane samoga raunalnoga jezikoslovlja. Npr. odgo-

nja kakvoe prijevoda ima jo mnogo prostora za napre-

varanje na pitanja (question answering, QA) postalo je

dak kod sustava za strojno prevoenje. Napredak se oe-

aktivnim podruje istraivanja za koje su izgraeni obi-

kuje u prilagodbi jezinih resursa odreenom podruju

ljeeni korpusi i organiziraju se posebna znanstvena na-

uporabe ovisno o temi ili korisniku, kao i u integraciji

tjecanja. Rjeavanje ovoga problema kree od pretrage

s postojeim sustavima za strojno potpomognuto pre-

temeljene na kljunim rijeima (na koju stroj obino od-

32

EN
BG
DE
CS
DA
EL
ES
ET
FI
FR
HU
IT
LT
LV
MT
NL
PL
PT
RO
SK
SL
SV

EN

61.3
53.6
58.4
57.6
59.5
60.0
52.0
49.3
64.0
48.0
61.0
51.8
54.0
72.1
56.9
60.8
60.7
60.8
60.8
61.0
58.5

BG
40.5

26.3
32.0
28.7
32.4
31.1
24.6
23.2
34.5
24.7
32.1
27.6
29.1
32.2
29.3
31.5
31.4
33.1
32.6
33.1
26.9

DE
46.8
38.7

42.6
44.1
43.1
42.7
37.3
36.0
45.1
34.3
44.3
33.9
35.0
37.2
46.9
40.2
42.9
38.5
39.4
37.9
41.0

CS
52.6
39.4
35.4

35.7
37.7
37.5
35.2
32.0
39.5
30.0
38.9
37.0
37.8
37.9
37.0
44.2
38.4
37.8
48.1
43.5
35.6

DA
50.0
39.6
43.1
43.6

44.5
44.4
37.8
37.9
47.4
33.0
45.8
36.8
38.5
38.9
45.4
42.1
42.8
40.3
41.0
42.6
46.6

EL
41.0
34.5
32.8
34.6
34.3

39.4
28.2
27.2
42.8
25.5
40.6
26.5
29.7
33.7
35.3
34.2
40.2
35.6
33.3
34.0
33.3

ES
55.2
46.9
47.1
48.9
47.5
54.0

40.4
39.7
60.9
34.1
26.9
21.1
8.0
48.7
49.7
46.2
60.7
50.4
46.2
47.0
46.6

ET
34.8
25.5
26.7
30.7
27.8
26.5
25.4

34.9
26.7
29.6
25.0
34.2
34.2
26.9
27.5
29.2
26.4
24.6
29.8
31.1
27.4

Markml Target language

FI FR HU IT LT LV
38.6 50.1 37.2 50.4 39.6 43.4
26.7 42.4 22.0 43.5 29.3 29.1
29.5 39.4 27.6 42.7 27.6 30.3
30.5 41.6 27.4 44.3 34.5 35.8
31.6 41.3 24.2 43.8 29.7 32.9
29.0 48.3 23.7 49.6 29.0 32.6
28.5 51.3 24.0 51.7 26.8 30.5
37.7 33.4 30.9 37.0 35.0 36.9
29.5 27.2 36.6 30.5 32.5
30.0 25.5 56.1 28.3 31.9
29.4 30.7 33.5 29.6 31.9
29.7 52.7 24.2 29.4 32.6
32.0 34.4 28.5 36.8 40.1
32.4 35.6 29.3 38.9 38.4
25.8 42.4 22.4 43.7 30.2 33.2
29.8 43.4 25.3 44.5 28.6 31.7
29.0 40.0 24.5 43.2 33.2 35.6
29.2 53.2 23.8 52.8 28.0 31.5
26.2 46.5 25.0 44.8 28.4 29.9
28.4 39.4 27.4 41.8 33.8 36.7
28.8 38.2 25.7 42.3 34.6 37.3
30.9 38.9 22.7 42.0 28.2 31.0

MT
39.8
25.9
19.8
26.3
21.1
23.8
24.6
20.5
19.4
25.3
18.1
24.6
22.2
23.3

22.0
27.9
24.8
28.7
28.5
30.0
23.7

NL
52.3
44.9
50.2
46.5
48.5
48.9
48.8
41.3
40.6
51.6
36.1
50.5
38.1
41.5
44.0

44.8
49.3
43.0
44.4
45.9
45.6

PL
49.2
35.1
30.2
39.2
34.3
34.2
33.9
32.0
28.8
35.7
29.8
35.2
31.6
34.4
37.1
32.0

34.5
35.8
39.0
38.2
32.2

PT
55.0
45.9
44.1
45.7
45.4
52.5
57.3
37.8
37.5
61.0
34.2
56.5
31.6
39.6
45.9
47.7
44.1

48.5
43.3
44.1
44.2

RO
49.0
36.8
30.7
36.5
33.9
37.2
38.1
28.0
26.5
43.8
25.7
39.3
29.3
31.0
38.9
33.0
38.2
39.4

35.3
35.8
32.7

SK
44.7
34.1
29.4
43.6
33.0
33.1
31.7
30.6
27.3
33.1
25.6
32.5
31.8
33.3
35.8
30.1
38.2
32.1
31.5

38.9
31.3

SL
50.7
34.1
31.4
41.3
36.2
36.3
33.9
32.9
28.2
35.6
28.2
34.7
35.3
37.1
40.0
34.6
39.8
34.4
35.1
42.6

33.5

SV
52.0
39.9
41.2
42.9
47.2
43.3
43.7
37.3
37.6
45.8
30.5
44.3
35.3
38.0
41.6
43.6
42.1
43.9
39.4
41.8
42.7

14: Kakvoa strojnoprevoditeljskoga prijevoda za sve parove izmeu 22 slubena jezika EU-a Machine translation between 22 EU-languages [34]
govara s itavim skupom potencijalno relevantnih do-

Ovaj je zadatak usko povezan sa zadatkom crpljenja

kumenata) prema scenariju u kojem korisnik postavlja

obavijesti (information extraction, IE), podrujem koje

konkretno pitanje, a sustav prua jedinstven odgovor,

je bilo iznimno popularno i utjecajno u vrijeme statis-

npr.:

tikoga prevrata u raunalnome jezikoslovlju, tj. u ra-

Pitanje: Koliko je godina imao Neil Armstrong kad je

stupio na Mjesec?
Odgoor: 38.

nim 1990-im. Cilj IE-a je pronai posebne obavijesti u

posebnim klasama dokumenata, a to bi moglo biti, npr.,
pronalaenje u novinskim lancima kljunih osoba koje
sudjeluju u preuzimanju tvrtki. Drugi mogui scena-

Premda se ovakva vrsta pretage nesumnjivo moe po-

rij, razraen iz izvjea o teroristikim napadima, tie se

vezati s ve spomenutim www-pretraivanjem, danas se

postupka s pomou kojega se iz teksta moe prepoznati

odgovaranje na pitanja smatra ponajprije sveobuhvat-

obrazac djelovanja, cilj, vrijeme ili mjesto napada i nje-

nim nazivom za prouavanja kakve sve vrste pitanja valja

gove posljedice. Popunjavanje takvih domenski ovisnih

razlikovati i kako se s njima valja postupati, kako se do-

obrazaca sredinja je osobina IE-a. Upravo zbog toga

kumente koji potencijalno sadre odgovor valja obrai-

IE predstavlja jo jedan primjer jezine tehnologije iza

vati i usporeivati (daju li suprotstavljene odgovore?), te

scene koja je sama jasno odijeljena od ostalih poddisci-

kako se odreena obavijest toan odgovor moe po-

plina (ili tehnologija), ali zbog praktinih razloga mora

uzdano crpiti iz dokumen(a)ta bez zanemarivanja kon-

biti ukljuena u ire uporabno okruje.

teksta.

33

Ovaj pristup zahtijeva stanovit oblik dubljega razumi-

Jezinotehnoloke aplikacije nerijetko

djeluju ispod poklopca i omoguuju uveanu
funkcionalnosti veih sustava.

jevanja teksta i stoga je manje robustan, a nije uope

mogu bez modula za generiranje teksta tj. novih reenica. Takav generator teksta u najveem broju sluajeva nije samostalna aplikacija ve je ukljuen u ira pro-

Godine 2009. Hrvatska izvjetajna novinska agen-

gramska okruja kao to su npr. medicinski informacij-

cija (HINA) [42] zapoela je s razvojem sustava za

ski sustavi gdje se podatci o podinim pacijentima skup-

(pred)obradbu svojih vijesti koja je ukljuila lematiza-

ljaju, spremaju, obrauju. Generiranje izvjea o stanju

ciju, prepoznavanje imena, klasikaciju vijesti prema za-

pacijenta samo je jedna od mnogih funkcionalnosti tak-

danoj shemi rubrika i crpljenje kljunih rijei. Ovaj su

vih sustava.

sustav zajedniki razvili Fakultet elektrotehnike i rau-

Niti jedna od tehnologija iz ova dva rubna podruja

narstva [43] i Filozofski fakultet, oba sastavnice Sveui-

jo ne postoje za hrvatski jezika osim nekoliko eksperi-

lita u Zagrebu.

menata koji su izvedeni za saimanje tekstova na hrvat-

Dva rubna podruja, koja katkada igraju ulogu samos-

skome jeziku [44] i generiranje teksta [45].

talnih, a katkada samo potpornih aplikacija, jesu saimanje teksta i generiranje teksta. Saimanje pokuava dati
sr duljega teksta u obliku kraega teksta, a postoji kao
ponuena funkcionalnost ve i u MS Wordu. Postupci
saimanja mahom su statistiki utemeljeni pri emu se
prvo identiciraju vane rijei u tekstu (npr. rijei koje
su visoko uestale u danome tekstu, ali su izrazito nisko
uestale u opoj jezinoj uporabi), a potom se pronalaze reenice u kojima se te rijei nalaze. Takve se reenice u dokumentu posebno obiljeavaju ili izdvajaju
ne bi li se od njih sastavio saetak. U tom najpopularnijem scenariju saimanje dokumenata vrsta je izdvajanja reenica: itav se tekst reducira na podskup svojih
reenica. Svi komercijalni sustavi za saimanje tekstova
koriste isti pristup. Alternativni pristup iskuava se u nekoliko istraivakih sredita i usmjeren je na sastavljanje
novih reenica, tj. na izgradnju saetka sastavljenog od

4.3 JEZINE TEHNOLOGIJE U

OBRAZOVANJU
Podruje jezinih tehnologija je visoko interdisciplinarno podruje koje ukljuuje strunjake iz jezikoslovlja, informacijskih znanosti, raunarskih znanosti, matematike, lozoje, psiholingvistike, kognitivne znanosti
i neuroznanosti itd. Kako se na Odsjeku za lingvistiku
Filozofskoga fakulteta Sveuilita u Zagrebu neprekinuto od 1950-ih prouavaju i pouavaju algebarskolingvistiki pristupi jezinome opisu, uvoenje posebnoga
smjera Raunalna lingvistika u dvogodinjem Diplomskome studiju lingvistike 2005. bilo je samo logian nastavak te tradicije. Slian je progam pokrenut na Sveuilitu u Zadru 2010.

reenica koje se ne moraju pojaviti u istome obliku u saimanome tekstu.

U veini tekstnih tehnologija istraivanja za

hrvatski jezik su manje razvijena od istraivanja
za druge europske jezike.

4.4 NACIONALNI PROJEKTI I

INICIJATIVE
Govornika hrvatskoga jezika ima oko 5,5 milijuna i taj
broj nipoto nije dovoljan da za odravanje skupoga razvoja novih jezinohtehnolokih proizvoda iskljuivo iz
komercijalnih izvora. Razvoj jezinih resursa i alata za

34

hrvatski jezik kota isto kao i za jezik s nekoliko stotina

vatska jezina riznica [54, 55] koja ukljuuje pisane teks-

milijuna govornika. Rezultat toga jest da je broj komer-

tove od 11. stoljea do dananih dana. Riznica je organi-

cijalno orijentiranih jezinotehnolokih tvrtki za hrvat-

zirana kao u tri glavna korpusa (starohrvatski, srednjehr-

ski jezik ravan nuli. Ulogu nancijskoga podupiratelja

vatski i suvremeni hrvatski) gdje se za prva dva rjeavaju

jezinotehnolokih aktivnosti djelomino preuzima dr-

bitni problemi dijakronijskih korpusa to u hrvatskome

ava, no sasvim sigurno ne u opsegu potrebnome za ra-

sluaju znai, transliteracija s tri razliita pisma (glago-

zvoj svih potrebnih jezinih resursa i alata.

ljice, irilice i latinice), rjeavanje nestandardnih pravopisnih rjeenja, individualne varijacije u uporabi pojedi-

Jeste li znali da se prva uporaba raunalnoga

usporednoga korpusa u kontrastivnoj lingvistici u
povijesti lingvistike dogodila u Zagrebu 1968?
U Hrvatskoj su se aktivnosti oko prikupljanja jezinih
resursa, tj. raunalnih korpusa, poele ve 1960-ih kad je
1967. eljko Bujas sastavio prvi hrvatski raunalni korpus i napravio njegovu konkordanciju [46] u Zavodu za
lingvistiku Filozofskoga fakulteta Sveuilita u Zagrebu.
Od tada je ta ustanova postala sredinjom ustanovom
u Hrvatskoj za istraivanja s podruja korpusne lingvistike. Godine 1968. u Zavodu se pod vodstvom Rudolfa
Filipovia po prvi puta u povijesti lingvistike uporabio usporedni raunalni korpus u kontrastivnolingvistikim istraivanjima [47]. Tijekom 1970-ih i 1980-ih
obavljala se raunalna obradba starih hrvatskih pisaca, a
sastavljanje Jednomilijunskoga korpusa hrvatskoga knjievnoga jezika zapoelo je 1976. pod vodstvom Milana
Mogua. Na temelju toga korpusa sastavljen je prvi hrvatski estotni rjenik [48].
Sastavljanje Hrvatskoga nacionalnoga korpusa [49] poelo je 1998. [50, 20], a opseg od 101 milijun rijei-

nih pismena itd.

Jeste li znali kako je najstariji hrvatski tiskani

rjenik Dictionarium quinque nobilissimarum
Europae linguarum Latinae, Italicae, Germanicae,
Dalmaticae et Ungaricae Fausta Vrania (1595)
ujedno i najstariji madarski tiskani rjenik?
Nakon istraivakih programa 1970-ih i 1980-ih, koji
su uobiajeno bili usmjereni prema raunalnoj obradbi
knjievnih tekstova, veinu istraivakih aktivnosti na
podruju raunalnoga jezikoslovlja, korpusnoga jezikoslovlja i jezinih tehnologija danas podupire Ministarstvo znanosti, obrazovanja i sporta kroz projekte povezane s jezinim tehnologijama. Jo je 1991. zapoeo
prvi takav projekt pod nazivom Raunalna obradba hrvatskoga knjievnoga jezika, 1996. je slijedio Raunalna
obradba hrvatskoga jezika, a 2002. Razitak hrvatskih jezinih resursa. Godine 2007. iz istoga su izvora poduprta tri osnovna istraivaka programa s po nekoliko
projekata usmjerenih na razvoj jezinih tehnologija za
hrvatski jezik:

pojavnica dohvatio je 2004. [51] Danas je najvei hrvat-

Raunalnolingvistiki modeli i jezine tehnologije

ski korpus hrWaC sastavljen na istome fakultetu 2011., a

za hrvatski jezik [56] gdje se sastavlja i odrava itav

obasie 1,2 milijarde rijei-pojavnica skupljenih s .hr in-

niz jezinih resursa i alata (npr. Hrvatski nacionalni

ternetske domene [52]. Od godine 2000. na istome se

korpus, Hrvatsko-engleski paralelni korpus, Hrvat-

fakultetu pod vodstvom Damira Borasa odvija opsena

ski morfoloki leksikon, Hrvatska ovisnosna banka

aktivnost digitalizacije starih hrvatskih jedno- i vieje-

stabala [57], Hrvatski wordnet [58], hibridni ozna-

zinih rjenika [53].

iva [59] i lematizator [15], ovisnosni parser za hr-

Pri Institutu za hrvatski jezik i jezikoslovlje 2004. zapo-

vatski, sustav za prepoznavanje imena i drugi alati za

elo je sastavljanje opsenoga korpusa pod nazivom Hr-

crpljenje obavijesti [60], itd.);

35

 Izvori za hrvatsku batinu i hrvatski europski identitet [61] s projektima koji se bave digitalizacijom starih hrvatskih rjenika i izradbom Hrvatskoga valencijskoga rjenika [62];

4.5 DOSTUPNOST ALATA I

RESURSA ZA HRVATSKI JEZIK
15 daje pregled trenutanoga stanja s jezinotehnolokom potporom hrvatskome jeziku. Ocjene postojeih

Hrvatska jezina riznica [54] gdje se niz projekata

alata i resursa temeljene su na uprosjeenoj procjeni ne-

bavi razliitim jezikoslovnim problemima zapoevi

koliko vodeih strunjaka s toga podruja koji su u mo-

od istraivanja hrvatskih narjeja i etimologije, do ra-

guem rasponu od 0 do 6 ocijenili stanje sluei se s ne-

zvoja semantikih mrea za izgradnju leksikih re-

koliko kriterija. Osnovni rezultati za hrvatske jezine

sursa. Svi ti projekti ukljuuju digitalizaciju skuplje-

tehnologije mogu se saeti u sljedeih nekoliko toaka:

nih jezinih podataka i izravno uveavaju broj dostupnih jezinih resursa za hrvatski jezik.

Kad je rije o veini temeljnih tehnolokih alata

i resursa (referentni korpusi, manji usporedni korpusi, veliki ektivni rjenici, opojavniitelji, MSDoznaivai, lematizatori, NERC sustavi itd.) hrvat-

Takoer na Sveuilitu u Rijeci projekt Govorne tehnologije [63] napravio je znaajan napredak u razvoju temeljnih resursa i alata za obradbu hrvatskoga govora kao
to su Hrvatski govorni korpus i prototipovi sustava za
ATR i TTS na hrvatskome.
Ovi su istraivaki programi otvorili mogunost da ra-

ski stoji relativno dobro.

Meutim, veliki sintaktiki obiljeeni korpusi nedostaju kao i veliki usporedni korpusi (npr. hrvatski prijevod pravne steevine EU). Mnogim postojeim resursima nedostaje standardiziran oblik, pa je
potrebna ozbiljna inicijativa da se standardiziraju ti
podatci kao i formati za razmjenu podataka.

zvoj jezinih tehnologija za hrvatski jezik uhvati korak s

Premda su u nekim podrujima ve zapoeli ekspe-

ostalim europskim jezicima, a istodobno su pruili pri-

rimenti, kao to su plitko parsanje (chunking), sa-

liku za ravnopravno sudjelovanje hrvatskih istraivakih

imanje, primjena ontolokih resursa, oni se odvi-

skupina u postojeim FP7 i ICT-PSP projektima s obzi-

jaju samo u akademskim krugovima, a dosegnuti re-

rom da je zadnji takav projekt (TELRI II) u kojem su

zultati su daleko od razine razvijenosti koju poka-

sudjelovali zavrio jo 2002.

zuju drugi europski jezici. Obradba multimedij-

Iz Republike Hrvatske Filozofski je fakultet Sveuilita

skih i multimodalnih dokumenata dobiva na va-

u Zagrebu bio je partnerom na projektu CLARIN, pot-

nosti, osobito digitalizacija u kontekstu ouvanja na-

hvatu koji nastoji oko izgradnje istraivake infrastruk-

cionalne kulturne batine, ali jezine tehnologije za

ture za istraivae s podruj humanistikih i drutve-

hrvatski jezik jo nisu ukljuene u te procese u do-

nih znanosti na razini itave Europe, a koja se infras-

voljnoj mjeri.

truktura temelji na jezinim resursima i alatima. Hrvat-

Na potpodrujima kao to su npr. sinteza govora,

ska je jedna od zemalja koje su izrazile spremnost pris-

prepoznavanje govora i crpljene obavijesti, postoje

tupiti CLARIN ERIC-u. Isti Fakultet je partner u FP7

pojedini prozvodi, ali ograniene ili visokospecijali-

projektu ACCURAT i ICT-PSP projektima LetsMT! i

zirane funkcional-nosti.

CESAR. Sveuilite u Zadru bilo je partnerom u ICTPSP projektu ATLAS.

Alati i resursi za naprednije jezine tehnologije kao

to su duboko parsanje, strojno prevoenje, teks-

36

tna semantika, obradba diskurza, generiranje jezika,

upravljanje dijalogom, itd. jednostavno za hrvatski
jo ne postoje.
Uzevi zajedno svu nancijsku potporu dobivenu kroz
spomenute projekte i programe s podruja jezinih tehnologija u rasponu od 2007. do 2012., moe se rei kako
je to jedva estina od stvarno potrebne potpore. Stoga
ne treba uditi kako se za jezine tehnologije za hrvatski jezik jo uvijek moe rei kako su u povojima. Broj
od oko 5,5 milijuna govornika hrvatskoga u Republici
Hrvatskoj i susjednim zemljama jednostavno je prema-

4. Sporadina razvijenost
5. Slaba ili nikakva razvijenost

Jezinotehnoloka razvijenost mjerena je prema sljedeim kriterijima:

Obradba govora: Kakvoa postojee tehnologije za
prepoznavanje govora, kakvoa postojee tehnologije za
sintezu govora, pokrivanje raznih podruja, broj i veliina postojeih govornih korpusa, broj i raznovrsnost
postojeih aplikacija govornih tehnologija.

len da bi se skup razvoj novih jezinotehnolokih pro-

Strojno prevoenje: Kakvoa postojeih tehnologija za

izvoda odravao samo trinim potrebama. Trenutano

strojno prevoenje, broj jezinih parova koji su zastup-

u Hrvatskoj nema tvrtke koja bi proizvodila jezinoteh-

ljeni, pokrivenost jezinih pojava i domena, kakvoa i

noloke alate jer se to ne smatra protabilnim. Stoga je

veliina postojeih usporedivih korpusa, broj i raznoli-

nastavak nancijske potpore iz javnih izvora kljuan, po-

kost postojeih primjena strojnoga prevoenja.

sebno imajui u vidu oekivani porast broja digitalnih

Analiza teksta: Kakvoa i zastupljenost postojeih teh-

dokumenata na hrvatskome s ukljuivanjem u Europsku

nologija za analizu teksta (morfologija, sintaksa, seman-

uniju 2013. kad e hrvatski jezik postati njezin 24. slu-

tika), pokrivenost razliitih jezinih pojava i podruja,

beni jezik.

broj i raznolikost postojeih primjena, kakvoa i opseg

postojeih (oznaenih) korpusa, kakvoa i pokrivenost

4.6 USPOREDBA IZMEU

JEZIKA

postojeih leksikih resursa (npr. Wordnet) i gramatika.

Jezini resursi: Kakvoa i opseg postojeih jednojezinih, govornih i usporednih korpusa, kakvoa i pokrive-

Trenutano stanje razvoja jezinih tehnologija znaajno

nost postojeih leksikih resursa i gramatika.

varira od jedne jezine zajednice do druge. Kako bi

Slike od 16 do 19 pokazuju kako je hrvatski za gotovo

se usporedilo stanje meu jezicima, ovo potpoglavlje

sve alate i resurse u skupini jezika koji su na dnu po ra-

predstavlja vrjednovanje temeljeno na dva ogledna po-

zvijenosti. Razvoj jezinih tehnologija za hrvatski uspo-

druja primjene jezinih tehnologija (strojno prevoe-

rediv je s ostalim jezicima maloga broja govornika kao

nje i obradba govora), jednom podruju primjene is-

to su estonski, letonski, litavski, slovaki, a u nekoj mjeri

pod poklopca (analiza teksta), kao i na temeljnim jezi-

danski i nski. Meutim, svi ovi jezici znatno zaostaju za

nim resursima potrebnim za izgradnju jezinotehnolo-

jezicima kao to su njemaki ili francuski, a niti za njih je-

kih aplikacija. Jezici su ocijenjeni prema skali od pet

zinotehnoloki resursi i alati ne doseu kakvou i opseg

bodova:

slinih resursa i alata koji su na raspolaganju za engleski

jezik. Stoga je upravo engleski jezik najnapredniji u go-

1. Izvrsna razvijenost

tovo svim podrujima premda i kod njega postoji znatan

2. Dobra razvijenost

broj manjkavosti u resursima koji bi se morali primijeniti

3. Umjerena razvijenost

u visoko kvalitetnim aplikacijama.

37

Koliina

Dostupnost

Kakvoa

Pokrivenost

Zrelost

Odrivost

Prilagodljivost

Prepoznavanje govora

Sinteza govora

Gramatika analiza

1.5

3.5

Semantika analiza

0.3

0.3

0.67

0.3

Generiranje teksta

Strojno prevoenje

Tekstovni korpus

2.5

Govorni korpusi

Usporedni korpusi

2.5

3.5

3.5

3.5

2.5

2.5

Jezini alati i aplikacije

Jezini resursi

Leksiki resursi
Gramatike

15: Stanje jezinih tehnologija za hrvatski jezik

4.7 ZAKLJUCI

temeljne tehnologije za analizu teksta i osnovni jezini

U oome nizu bijelih knjiga po prvi se puta za 30 europ-

ali je primjena npr. semantikih metoda jo uvijek da-

skih jezika pokualo procijeniti njihovu jezinotehnoloku

leko. Stoga je potreban irok zajedniki napor kako bi se

razvijenost i dati njihovu meusobnu usporedbu. Uoa-

postigao ambiciozni cilj uspostave visoke razine razvije-

vanjem manjkaosti, potreba i nedostataka, europska je-

nosti jezinih tehnologija za sve europske jezike vidljiv

zinotehnoloka zajednica i zainteresirani dionici sad su

u, npr. strojnome prevoenju visoke kakvoe.

u poloaju sastaviti opsean program istraivanja i ra-

Ne moemo zaista biti optimistini glede jezinih teh-

zoja usmjeren na izgradnju istinski viejezine i tehno-

nologija za hrvatski jezik. Na tom podruju u Hrvatskoj

loko potpomognute komunikacije unutar cijele Europe.

postoji istraivaka scena u nastajanju, ponajprije na sve-

Rezultati ovoga niza bijelih knjiga pokazuju kako pos-

uilitima i u istraivakim institutima, ali mala ili sred-

toje znaajne razlike u razvijenosti jezinih tehnologija

nja poduzea, kao potencijalni korisnici ili proizvoai

za razliite europske jezike. Dok za neke jezike postoje

jezinih tehnologija za hrvatski jezik, gotovo ne postoje.

dobre aplikacije visoke kakvoe i slobodno dostupni je-

Razne su ustanove uloile napore na istraivanje i ra-

zini resursi, za drugi, obino maloljudniji jezici, poka-

zvoj jezinotehnolokih proizvoda kao to su veliki hr-

zuju znaajne manjkavosti. Mnogim jezicima nedostaju

vatski korpusi, obradba morfologije, strojno prevoenje,

resursi. Drugi pak jezici imaju temeljne alate i resurse,

38

obradba govora, itd. No ti se alati i resursi moraju dalje

gija u osiguravanju odrivoga razvoja hrvatskoga jezika u

razvijati, a za to je potrebna potpora. Prema procjenama

21. stoljeu i u izazovima koje e pred njega staviti uloga

danim podrobno u ovome izvjeu, potrebna je urna i

jednoga od slubenih jezika EU, jo uvijek nije pokre-

neposredna akcija kako bi se osigurala daljnja nova pos-

nuta nikakva opsena inicijativa na nacionalnoj razini

tignua za hrvatski jezik. Sasvim je razvidno kako se mo-

koja bi skrbila o stvaranju velikih resursa, alata i servisa za

raju pojaati napori u stvaranju jezinih resursa za hrvat-

hrvatski jezik, o partnerstvu izmeu vlade, istraivanja i

ski jezik i openito poduprijeti njihovo istraivanje, ino-

gospodarstva ne bi li se razvio struno-komercijalni klas-

vacije i razvoj. Potreba za velikim koliinama podataka

ter za hrvatske jezine tehnologije. Vjerujemo kako bi

kao i krajnja sloenost jezinotehnolokih sustava uka-

ta inicijativa morala imati institucionalni okvir u obliku

zuje na potrebu razvoja kljune nove infrastrukture koja

posebnoga sredita kompetencija/izvrsnosti koji bi mo-

e omoguiti suradnju i dijeljenje resursa, alata i znanja.

gao dobiti potporu iz strukturnih fondova EU s ciljem

Javna nancijska potpora jezinim tehnologijama u

poticanja poslovno orijentiranih istraivanja, promica-

Europi je relativno niska kad ju se usporedi s trokovima

nja suradnje unutar podruja izmeu tvrtki i istraiva-

prevoenja i viejezinoga pristupa u SAD-u [64]. U

kih ustanova na razvoju novih proizvoda i tehnologija,

Hrvatskoj je javno nanciranje razvoja jezinih tehno-

te podizanja kompetitivnost hrvatskih tvrtki na tritu

logija jo i manje nego u mnogim usporedivim europ-

EU kojega e Hrvatska postati sastavni dio ve 2013.

skim zemljama, ukljuujui i susjedne zemlje poput Slo-

Dugoroni je cilj META-NET-a omoguiti stvaranje vi-

venije ili Madarske. Nerijetko postoji nedostatak kon-

sokokvalitetnih jezinih tehnologija za sve jezike. To

tinuiteta u nancijskoj potpori istraivanjima i razvoju.

zahtijeva da svi dionici u tome procesu politiari, is-

Kratkoroni projekti ili programi smjenjuju se s razdob-

traivai, poduzetnici ujedine svoje napore. Rezultat

ljima slabije ili nikakve potpore. Uz to postoji i opi ne-

e biti tehnologije koje e omoguiti nadilaenje posto-

dostatak koordinacije s programima u ostalim zemljama

jeih prepreka i izgradnju mostova izmeu europskih je-

EU-a na razini Europske komsije. Premda postoji go-

zika, pripremajui put za politiko i ekonomsko jedins-

rua potreba prepoznavanja vanosti jezinih tehnolo-

tvo kroz kulturnu raznolikost.

39

Izvrsna
podrka

Dobra
podrka
Engleski

Djelomina
podrka
Finski
Francuski
Nizozemski
Talijanski
Portugalski
panjolski
eki
Ruski

Sporadina
podrka
Baskijski
Bugarski
Danski
Estonski
Galicijski
Grki
Irski
Katalonski
Norveki
Poljski
Srpski
Slovaki
Slovenski
vedski
Maarski

Slaba podrka/
odsutnost podrke
Islandski
Hrvatski
Latvijski
Litavski
Malteki
Rumunjski

16: Obrada govora: stanje jezinih tehnologija za 30 slubenih jezika Europe

Izvrsna
podrka

Dobra
podrka
Engleski

Djelomina
podrka
Francuski
panjolski

Sporadina
podrka
Nizozemski
Talijanski
Katalonski
Poljski
Rumunjski
Maarski
Ruski

Slaba podrka/
odsutnost podrke
Baskijski
Bugarski
Danski
Estonski
Finski
Galicijski
Grki
Irski
Islandski
Hrvatski
Latvijski
Litavski
Malteki
Norveki
Portugalski
Srpski
Slovaki
Slovenski
vedski
eki

17: Strojno prevoenje: stanje jezinih tehnologija za 30 slubenih jezika Europe

40

Izvrsna
podrka

Dobra
podrka
Engleski

Djelomina
podrka
Francuski
Nizozemski
Talijanski
panjolski
Ruski

Sporadina
podrka
Baskijski
Bugarski
Danski
Finski
Galicijski
Grki
Katalonski
Norveki
Poljski
Portugalski
Rumunjski
Slovaki
Slovenski
vedski
eki
Maarski

Slaba podrka/
odsutnost podrke
Estonski
Irski
Islandski
Hrvatski
Latvijski
Litavski
Malteki
Srpski

18: Analiza teksta: stanje jezinih tehnologija za 30 slubenih jezika Europe

Izvrsna
podrka

Dobra
podrka
Engleski

Djelomina
podrka
Francuski
Nizozemski
Talijanski
Poljski
panjolski
vedski
eki
Maarski
Ruski

Sporadina
podrka
Baskijski
Bugarski
Danski
Estonski
Finski
Galicijski
Grki
Katalonski
Hrvatski
Norveki
Portugalski
Rumunjski
Srpski
Slovaki
Slovenski

Slaba podrka/
odsutnost podrke
Irski
Islandski
Latvijski
Litavski
Malteki

19: Jezini resursi: stanje jezinih tehnologija za 30 slubenih jezika Europe

41

5
O META-NET-U
META-NET je mrea izvrsnosti koju podupire Europ-

koherentnu i kohezivnu jezinotehnoloku zajednicu u

ska komisija. Mrea se trenutano sastoji od 54 lana

Europi ukljuivanjem predstavnika iz sasvim razliitih i

iz 33 europske zemlje [65]. META-NET organizira

rascjepkanih skupina dionika. Ova bijela knjiga prire-

META (Multilingual Europe Technology Alliance),

ena je zajedniki za jo 29 jezika. Zajednika tehno-

rastuu zajednicu europskih profesionalaca i organiza-

loka vizija u tri podrune skupine za META-VISION.

cija u podruju jezinih tehnologija.

Uspostavljeon je META tehnoloko vijee kako bi se

META-NET skrbi o tehnolokim temeljima za istinsko

raspravio SRA na temelju iroke rasprave u itavoj jezi-

viejezino europsko drutvo koje e:

notehnolokoj zajednici.

omoguiti komunikaciju i suradnju meu jezicima;

za sve Europljane osigurati jednak pristup informacijama i znanju na bilo kojem jeziku;
doraivati i unaprjeivati funkcionalnosti umreene
informacijske tehnologije.

META-SHARE stvara otvorenu i distribuiranu platforma za razmjenu i dijeljenje resursa.

Peer-to-peer

mrea digitalnih repozitorija sadravat e jezine podatke, alate i web servise koji su dokumentirani visokokvalitetnim metapodatcima i organizirani u standardizirane kategorije. Resursi su odmah dostupni i jed-

Ova mrea izvrsnosti podupire Europu koja se ujedi-

noobrazno pretraivi. Raspoloivi resursi ukljuuju bes-

njuje u jedinstveno digitalno trite i jedinstven infor-

platnu i otvorenu grau kao i ograniene, komercijalno

macijski prostor. Ona potie i promie viejezine teh-

dostupne, naplative resurse.

nologije za sve europske jezike. Te tehnologije podu-

META-RESEARCH izgrauje mostove prema susjed-

piru strojno prevoenje, automatsko generiranje sadr-

nim istraivakim i tehnolokim podrujima. Ove ak-

aja, obradbu obavijesti i upravljanje znanjem u veli-

tivnosti trae napredak u drugim podrujima s kojih bi

kome broju podruja i primjena. Te tehnologije takoer

inovativna istraivanja mogli unaprijediti jezine tehno-

omoguuju intuitivna, jezino utemeljena suelja u ras-

logije. Aktivnosti se osobito usredotouju na izvoe-

ponu od kuanskih ureaja, strojeva i vozila do raunala

nje vrhunskih istraivanja u podruju strojnoga prevo-

i robota. S poetkom od 1. veljae 2010., META-NET

enja, prikupljanja podataka, prireivanja skupova po-

je ve organizirao mnogobrojne aktivnosti u tri osnovna

dataka i organizacije jezinih resursa za potrebe vrjed-

smjera djelovanja: META-VISION, META-SHARE i

novanja; potom u podruju katalogiziranja jezinoteh-

META-RESEARCH.

nolokih alata i metoda; u organizaciji radionica i raznih

META-VISION skrbi o zajednici dinaminih i utje-

oblika dodatne izobrazbe lanova jezinotehnoloke za-

cajnih dionika koja je okupljena oko jedinstvene vizije

jednice.

i zajednikoga istraivakoga plana (Strategic Research

Agenda, SRA). Glavni cilj ovih aktivnosti jest izgraditi

oce@meta-net.eu http://www.meta-net.eu

42

1
EXECUTIVE SUMMARY
Information technology changes our everyday lives. We

projected that from mid 2013 the Croatian language

typically use computers for writing, editing, calculating,

will become the 24th ocial language of the European

and information searching, and increasingly for reading,

Union. Starting with that moment it will be expected

listening to music, viewing photos and watching movies.

that for Croatian the whole range of dierent language

We carry small computers in our pockets and use them

resources, tools and services will be accessible, similar to

to make phone calls, write emails, get information and

the ones that already exist and are being developed fur-

entertain ourselves, wherever we are. How does this

ther for other EU languages. Search engines providing

massive digitisation of information, knowledge and ev-

full-text search with all word forms in which Croatian

eryday communication aect our language? Will our

words could appear, dictation systems, i. e., speech to

language change or even disappear? What are the Croa-

text systems for Croatian, or maybe the most impor-

tian languages chances of survival?

tant machine translation systems to and from Croa-

Many of the worlds 6,000 languages will not survive in

tian, are just some of examples of important language

a globalised digital information society. It is estimated

technologies. ese systems are not expected as research

that at least 2,000 languages are doomed to extinction

prototypes only, but also as useful commercial prod-

in the decades ahead. Others will continue to play a role

ucts. We cant expect that they will be developed for the

in families and neighbourhoods, but not in the wider

Croatian language by researchers dealing with English,

business and academic world. e status of a language

French, German, Czech, Slovenian or Serbian, but we

depends not only on the number of speakers or books,

have to develop these language resources, tools and ser-

lms and TV stations that use it, but also on the pres-

vices on our own. However, this will be easier to achieve

ence of the language in the digital information space and

if we harmonise and coordinate our eorts with similar

soware applications.

eorts for other EU languages. It is exactly what the ini-

In todays information society accessibility of informa-

tiative described in this publication is about.

tion in your mother tongue is considered to be the civilisational level necessary for overcoming the digital di-

is white paper for the Croatian language demon-

vide. e linguistic communities without developed

strates that a basic language research environment exists

language technologies for their language will remain on

in Croatia, although the language industry is not really

the other side of digital divide. When it comes to the

developed. Despite the fact that a small number of tech-

Croatian language and its language technologies, it is

nologies and resources for Croatian exist, there are fewer

not just the assurance that it will be able to participate

of them developed for the Croatian language than for

on equal grounds with other languages in our globalised

other Slavic languages, e. g., Czech, and far fewer than

information society, but even more it is about the im-

for the major EU languages, like English, German or

minent change of its sociolinguistic conditions. It is

French.

43

Although in Croatia theres a half-century long tradi-

According to the assessment detailed in this report, fo-

tion of research in computational linguistics, natural

cused action must be taken in order to bring the Croa-

language processing and corpus linguistics (with com-

tian language resources and tools at the level of qual-

piling such important language resources as the Croat-

ity and quantity of language resources and tools that al-

ian Frequency Dictionary, the Croatian National Cor-

ready exist for other European languages.

pus, the Croatian-English Parallel Corpus, the Croatian Morphological Lexicon, the Croatian Dependency
Treebank, etc.), it cant be assumed that the current status of language technologies is satisfactory. Beside the
nationally funded projects unfortunately, still only
few of them since 2008 started more substantial funding through ve EC projects: CLARIN, ACCURAT,
LetsMT!, ATLAS, XLike; but they are also mainly ori-

META-NETs vision is high-quality language technology for all languages that supports political and economic unity through cultural diversity. is technology
will help tear down existing barriers and build bridges
between Europes languages. is requires all stakeholders in politics, research, business, and society to unite
their eorts for the future.

ented towards solving individual problems or providing

is white paper series complements the other strategic

technological solutions, and rarely towards advancing

actions taken by META-NET. Up-to-date information

the overall situation of language technologies for Croa-

such as the current version of the META-NET vision

tian. For the Croatian language the sixth project CE-

paper [2] or the Strategic Research Agenda (SRA) can

SAR takes exactly this role within the wider META-

be found on the META-NET web site: http://www.

NET initiative, by producing this white paper.

meta-net.eu.

44

2
LANGUAGES AT RISK: A CHALLENGE FOR
LANGUAGE TECHNOLOGY
We are witnesses to a digital revolution that is dramat-

the creation of dierent media like newspapers, ra-

ically impacting communication and society. Recent

dio, television, books, and other formats satised

developments in digitised and network communication

dierent communication needs.

technology are sometimes compared to Gutenbergs invention of the printing press. What can this analogy tell

In the past twenty years, information technology helped

us about the future of the European information society

to automate and facilitate many processes:

and our languages in particular?

desktop publishing soware replaces typewriting
and typesetting;

The digital revolution is comparable to

Gutenbergs invention of the printing press.

Microso PowerPoint replaces overhead projector

transparencies;
e-mail allows documents to be sent and received

Aer Gutenbergs invention, real breakthroughs in

communication and knowledge exchange were accomplished by eorts such as Luthers translation of the
Bible into vernacular language. In subsequent centuries,
cultural techniques have been developed to better handle language processing and knowledge exchange:

more quickly than using a fax machine;

Skype oers cheap Internet phone calls and hosts
virtual meetings;
audio and video encoding formats make it easy to exchange multimedia content;
web search engines provide keyword-based access;

the orthographic and grammatical standardisation

of major languages enabled the rapid dissemination
of new scientic and intellectual ideas;
the development of ocial languages made it possible for citizens to communicate within certain (often political) boundaries;
the teaching and translation of languages enabled exchanges across languages;
the creation of editorial and bibliographic guidelines
assured the quality of printed material;

online services like Google Translate produce quick,

approximate translations;
social media platforms such as Facebook, Twitter
and Google+ facilitate communicaton, collaboration and information sharing.
Although such tools and applications are helpful, they
are not yet capable of supporting a fully-sustainable,
multilingual European society in which information
and goods can ow freely.

45

2.1 LANGUAGE BORDERS

HOLD BACK THE EUROPEAN
INFORMATION SOCIETY

this ubiquitous digital linguistic divide has not gained

We cannot predict exactly what the future information

are doomed to disappear?

much public attention; yet, it raises a very pressing question: Which European languages will thrive in the networked information and knowledge society, and which

society will look like. However, there is a strong likelihood that the revolution in communication technology is bringing together people who speak dierent languages in new ways. is is putting pressure both on individuals to learn new languages and especially on developers to create new technology applications to ensure
mutual understanding and access to shareable knowledge. In the global economic and information space,
there is increasing interaction between dierent languages, speakers and content thanks to new types of media. e current popularity of social media (Wikipedia,
Facebook, Twitter, YouTube, and, recently, Google+) is
only the tip of the iceberg.

2.2 OUR LANGUAGES AT RISK

While the printing press helped step up the exchange of
information in Europe, it also led to the extinction of
many European languages. Regional and minority languages were rarely printed and languages such as Cornish and Dalmatian were limited to oral forms of transmission, which in turn restricted their scope of use. Will
the Internet have the same impact on our modern languages?
Europes approximately 80 languages are one of our richest and most important cultural assets, and a vital part

The global economy and information

space confronts us with dierent languages,
speakers and content.

of this unique social model [4]. While languages such

as English and Spanish are likely to survive in the emerging digital marketplace, many European languages could
become irrelevant in a networked society. is would

Today, we can transmit gigabytes of text around the

weaken Europes global standing, and run counter to the

world in a few seconds before we recognise that it is in

strategic goal of ensuring equal participation for every

a language that we do not understand. According to a

European citizen regardless of language.

recent report from the European Commission, 57% of

Internet users in Europe purchase goods and services in
non-native languages; English is the most common foreign language followed by French, German and Spanish.
55% of users read content in a foreign language while
35% use another language to write e-mails or post com-

The variety of languages in Europe is one of its

richest and most important cultural assets.

ments on the Web [3]. A few years ago, English might

have been the lingua franca of the Web the vast majority of content on the Web was in English but the

According to a UNESCO report on multilingualism,

situation has now drastically changed. e amount of

languages are an essential medium for the enjoyment of

online content in other European (as well as Asian and

fundamental rights, such as political expression, educa-

Middle Eastern) languages has exploded. Surprisingly,

tion and participation in society [5].

46

2.3 LANGUAGE TECHNOLOGY

IS A KEY ENABLING
TECHNOLOGY
In the past, investments in language preservation fo-

be able to achieve a really eective interactive, multimedia and multilingual user experience in the near future.

Europe needs robust and aordable language

technology for all European languages.

cussed primarily on language education and translation. According to one estimate, the European market
for translation, interpretation, soware localisation and
website globalisation was 8.4 billion in 2008 and is
expected to grow by 10% per annum [6]. Yet this gure covers just a small proportion of current and future

2.4 OPPORTUNITIES FOR

LANGUAGE TECHNOLOGY

needs in communicating between languages. e most

In the world of print, the technology breakthrough was

compelling solution for ensuring the breadth and depth

the rapid duplication of an image of a text using a suit-

of language usage in Europe tomorrow is to use appro-

ably powered printing press. Human beings had to do

priate technology, just as we use technology to solve our

the hard work of looking up, assessing, translating, and

transport and energy needs among others.

summarising knowledge. We had to wait until Edison

Language technology targeting all forms of written text

to record spoken language and again his technology

and spoken discourse can help people to collaborate,

simply made analogue copies.

conduct business, share knowledge and participate in

Language technology can now simplify and automate

social and political debate regardless of language barri-

the processes of translation, content production, and

ers and computer skills. It oen operates invisibly inside

knowledge management for all European languages. It

complex soware systems to help us already today to:

can also empower intuitive speech-based interfaces for

household electronics, machinery, vehicles, computers

nd information with a search engine;

and robots. Real-world commercial and industrial ap-

check spelling and grammar in a word processor;

plications are still in the early stages of development,

view product recommendations in an online shop;

yet R&D achievements are creating a genuine window

follow the spoken directions of a navigation system;

of opportunity. For example, machine translation is al-

translate web pages via an online service.

ready reasonably accurate in specic domains, and experimental applications provide multilingual informa-

Language technology consists of a number of core ap-

tion and knowledge management, as well as content

plications that enable processes within a larger applica-

production, in many European languages.

tion framework. e purpose of the META-NET lan-

As with most technologies, the rst language applica-

guage white papers is to focus on how ready these core

tions such as voice-based user interfaces and dialogue

enabling technologies are for each European language.

systems were developed for specialised domains, and of-

To maintain our position in the frontline of global inno-

ten exhibit limited performance. However, there are

vation, Europe will need language technology, tailored

huge market opportunities in the education and enter-

to all European languages, that is robust and aordable

tainment industries for integrating language technolo-

and can be tightly integrated within key soware envi-

gies into games, edutainment packages, libraries, simu-

ronments. Without language technology, we will not

lation environments and training programmes. Mobile

47

information services, computer-assisted language learning soware, eLearning environments, self-assessment

tools and plagiarism detection soware are just some

2.5 CHALLENGES FACING

LANGUAGE TECHNOLOGY

of the application areas in which language technology

Although language technology has made considerable

can play an important role. e popularity of social

progress in the last few years, the current pace of tech-

media applications like Twitter and Facebook suggest a

nological progress and product innovation is too slow.

need for sophisticated language technologies that can

Widely-used technologies such as the spelling and gram-

monitor posts, summarise discussions, suggest opinion

mar correctors in word processors are typically mono-

trends, detect emotional responses, identify copyright

lingual and are only available for a handful of languages.

infringements or track misuse.

Technological progress needs to be accelerated.

Online machine translation services, although useful

Language technology helps overcome

the disability of linguistic diversity.

for quickly generating a reasonable approximation of a

documents contents, are fraught with diculties when
highly accurate and complete translations are required.
Due to the complexity of human language, modelling

Language technology represents a tremendous opportu-

our tongues in soware and testing them in the real

nity for the European Union. It can help to address the

world is a long, costly business that requires sustained

complex issue of multilingualism in Europe the fact

funding commitments. Europe must therefore main-

that dierent languages coexist naturally in European

tain its pioneering role in facing the technological chal-

businesses, organisations and schools. However, citi-

lenges of a multiple-language community by inventing

zens need to communicate across the language borders

new methods to accelerate development right across the

of the European Common Market, and language tech-

map. ese could include both computational advances

nology can help overcome this nal barrier, while sup-

and techniques such as crowd sourcing.

porting the free and open use of individual languages.

for our global partners when they begin to support

2.6 LANGUAGE ACQUISITION

IN HUMANS AND MACHINES

their own multilingual communities. Language tech-

To illustrate how computers handle language and why it

nology can be seen as a form of assistive technology

is dicult to program them to process dierent tongues,

that helps overcome the disability of linguistic diver-

lets look briey at the way humans acquire rst and sec-

sity and makes language communities more accessible to

ond languages, and then see how language technology

each other. Finally, one active eld of research is the use

systems work.

of language technology for rescue operations in disas-

Humans acquire language skills in two dierent ways.

ter areas, where performance can be a matter of life and

Babies acquire a language by listening to the real inter-

death: Future intelligent robots with cross-lingual lan-

action between their parents, siblings and other family

guage capabilities have the potential to save lives.

members. From the age of about two, children produce

Looking even further ahead, innovative European multilingual language technology will provide a benchmark

48

their rst words and short phrases. is is only possi-

systems. Experts in the elds of linguistics, computa-

ble because humans have a genetic disposition to imitate

tional linguistics and computer science rst have to en-

and then rationalise what they hear.

code grammatical analyses (translation rules) and com-

Learning a second language at an older age requires

pile vocabulary lists (lexicons). is is very time con-

more cognitive eort, largely because a child is not im-

suming and labour intensive. Some of the leading rule

mersed in a language community of native speakers. At

based machine translation systems have been under con-

school, foreign languages are usually acquired by learn-

stant development for more than 20 years. e great

ing grammatical structure, vocabulary and spelling using

advantage of rule-based systems is that the experts have

drills that describe linguistic knowledge in terms of ab-

more detailed control over the language processing.

stract rules, tables and examples.

is makes it possible to systematically correct mistakes

in the soware and give detailed feedback to the user, es-

Humans acquire language skills in two dierent

ways: learning from examples and learning the
underlying language rules.

pecially when rule-based systems are used for language

learning. However, due to the high cost of this work,
rule-based language technology has so far only been developed for a few major languages.

Moving now to language technology, the two main

types of systems acquire language capabilities in a similar manner. Statistical (or data-driven) approaches obtain linguistic knowledge from vast collections of con-

The two main types of language technology

systems acquire language in a similar manner.

crete example texts. While it is sucient to use text in a

single language for training, e. g., a spell checker, paral-

As the strengths and weaknesses of statistical and rule

lel texts in two (or more) languages have to be available

based systems tend to be complementary, current re-

for training a machine translation system. e machine

search focussed on hybrid approaches that combine the

learning algorithm then learns patterns of how words,

two methodologies. However, these approaches have so

short phrases and complete sentences are translated.

far been less successful in industrial applications than in

is statistical approach usually requires millions of sen-

the research lab.

tences to boost performance quality. is is one rea-

As we have seen in this chapter, many applications

son why search engine providers are eager to collect as

widely used in todays information society rely heavily

much written material as possible. Spelling correction

on language technology, particularly in Europes eco-

in word processors, and services such as Google Search

nomic and information space. Although this technol-

and Google Translate, all rely on statistical approaches.

ogy has made considerable progress in the last few years,

e great advantage of statistics is that the machine

there is still huge potential to improve the quality of lan-

learns quickly in a continuous series of training cycles,

guage technology systems. In the next section, we de-

even though quality can vary randomly.

scribe the role of Croatian in the European information

e second approach to language technology, and to

society and assess the current state of language technol-

machine translation in particular, is to build rule-based

ogy for the Croatian language.

49

3
THE CROATIAN LANGUAGE IN THE
EUROPEAN INFORMATION SOCIETY
3.1 GENERAL FACTS
e Croatian language belongs to the West-South Slavic
subgroup of the Slavic branch of the Indo-European linguistic family. Currently over 5.5 million people speak
Croatian as their native language. e Croatian language consists of the dialects and standard national language of the Croats, which is the ocial language of
more than 4 million people in the Republic of Croatia
and is, along with Bosnian and Serbian, one of three ofcial languages in Bosnia and Herzegovina, where it is
spoken by about 700,000 people. However, the Croatian language is also spoken by members of national minorities in Croatia as well as by autochthonous Croatian ethnic and linguistic minorities in Serbia, Montenegro, Slovenia, Hungary, Austria, Slovakia and Italy, who
either reside upon territories of former Croatian lands
or emigrated due to historically conditioned exoduses
throughout the centuries.

diaspora is located in Germany, followed by the USA,

Canada and Australia, and they also occasionally use
the Croatian language. eir active use of the Croatian
language mainly depends on the generation of emigration they belong to. However, in many countries, especially in Europe, there are additional school programs in
Croatian organized and nanced by the Croatian government.
e ocial status of the Croatian language in Croatia
is dened by the Constitution of the Republic of Croatia. According to Article 12 of the Constitution: e
Croatian language and the Latin script shall be in ocial
use in the Republic of Croatia. In individual local units,
another language and Cyrillics or some other script may
be introduced into ocial use together with the Croatian language and Latin script under conditions specied
by law. Since Croatia is expected to join the European
Union in 2013, the Croatian language will then become
the 24th ocial language of the EU.
In Croatia there is still not a unied language law stip-

Croatian is the language of government and

administration, all levels of the school system, and
the language of business and general day-to-day
interactions in Croatia.

ulating the usage of Croatian as an ocial language

in public matters. Eorts to introduce a language act
have been undertaken on a few occasions since Croatia
gained independence, but so far none of them succeeded
in gaining the support of the Croatian Government and

Due to intensive economically and politically condi-

did not enter parliamentary procedure. e last attempt

tioned emigration aer the two World Wars in the 20th

was made in April 2010. However, certain articles re-

century, Croatian is also spoken within the Croatian

garding the usage of Croatian as an ocial state lan-

linguistic community in a number of other European

guage in ocial matters are found within acts on edu-

countries and overseas. e largest Croatian economic

cation, court procedures etc. So far, legislation states no

50

requirement for a compulsory test or examination as a

Croatia has a sizable diaspora that oen still speaks

prerequisite for naturalization. e Citizenship Act [7]

Croatian (see Figure 1). Croatian ethnic and linguistic

presupposes that a foreign person applying for Croatian

minorities live in many European countries due to his-

citizenship is familiar with the Croatian language and

torical migrations beginning from the 16th century, as

alphabet.

well as recent, mostly economical and political emigra-

According to the 2001 census, Croatia has 4,437,460

residents, of whom 89.63% are Croats. Serbs are the
most signicant national minority, comprising 4.54%
of the population, while each remaining national minority makes up less than 0.5% of the population:
the Bosniaks (0.47%), Albanians (0.34%), Slovenians
(0.30%), Montenegrins (0.11%) and others in less signicant numbers. Croatian is the native language of
96% of all residents. National minorities declared to
speak these languages: Albanian, Bosnian, Bulgarian,
Czech, Hebrew, Hungarian, German, Istro-Romanian,
Italian, Macedonian, Montenegrin, Polish, Roma, Romanian, Russian, Rusyn, Slovak, Serbian, Turkish and
Ukrainian. Four minority languages, Serbian, Hungarian, Italian and Czech, have earned the right to the ofcial use of their minority language and script in certain districts according to their share in the population,
which must amount to 1/3 of the general population in
a local government district. As of 2009, there are 27
districts in Croatia where national minorities have the
right to the ocial use of their language in local administration. at right is used to a high degree in Istarska
County, where Italian is the native language of 20,521
residents, but bilingual street signs can be found even in
areas where there is no Italian minority. e Republic
of Croatia ratied the European Charter for Regional
or Minority Languages in 1997.

tion. e most numerous groups are the so-called Burgenland Croats in Austria (presumably about 50,000),
and about the same number of Croats live in Hungary.
In Austria, the Croats actively use Burgenland Croatian.
is variant of Croatian, which has been standardized
according to somewhat dierent rules than standard
Croatian, is one of Austrias ocial minority languages.
ere are a number of kindergartens and schools in Burgenland that use Burgenland Croatian. On the other
hand, the Croatian standard language is the ocial minority language in Hungary. In Italy at the moment live
about 3,000 Croats, who use a variant of Croatian called
Molise Croatian that is also taught in schools in three
communities inhabited by Croats in Molise. e number of Croats in Serbia, specically in the province of
Vojvodina where Croats are a national minority, is difcult to establish, since a number of ethnic Croats are
declared as so-called Bunjevci, mostly for political reasons. Although many Croats were expelled from Serbia
aer Croatia gained independence from Yugoslavia, it
is assumed that there are still more than 100,000 Croats
in Serbia. In other European countries, a Croatian autochthonous minority lives in Montenegro (7,000 to
10,000), the Czech Republic (less than 1,000), Slovakia
(4,000) and Romania (7,500). e number of Croats
in Slovenia is about 50,000, but only a small number of
them are an autochthonous minority, mostly in settlements along the border, and more of them are recent

e recently conducted 2011 census, which was carried

economic emigrants. So, as a minority language, Croa-

out according to international statistical standards, and

tian is an ocial minority language in Serbia (as one of

thus enumerated all citizens of the Republic of Croatia,

seven ocial languages in the province of Vojvodina),

foreign citizens and stateless persons who reside in the

Montenegro, Austria and Hungary, and in Italy, Molise

Republic of Croatia, has not yet provided ocial gures

Croatian is recognized as a linguistic minority.

on language usage.

51

1: Croats in neighbouring states [8]

52

3.2 CROATIAN DIALECTS

wider area, but at present the tokavian dialect prevails.

e dialectal picture of Croatia is composed of three

as far North as the rivers Kupa and Sava, and as far east

dialectal groups: akavian, Kajkavian and tokavian

as the Una-Dinara-Cetina line. Aer migrations, aka-

(see Figure 2). Dialects belonging to all three dialec-

vian dialects were ousted mostly to the coastal regions

tal groups are spoken throughout the Republic of Croa-

and islands, while the akavian dialects inland began

tia. All Croatian dialects belong to the Central South

to dier according to the degree of tokavian inuence.

Slavic diasystem of the Slavic linguistic branch, and on

e Kajkavian dialects were also once spoken much fur-

the South-Slavic territory it comprises part of the di-

ther to the East, where the tokavian prevails today.

alectal continuum between the Slovenian type in the

e akavian, Kajkavian and tokavian dialectal groups

North-West and the Macedonian-Bulgarian type in the

dier on all linguistic levels: phonological, morpholog-

South-East. e names of those dialectal groups are

ical, syntactic and lexical, and each level includes a num-

based upon the use of the interrogative pronouns a,

ber of archaisms and innovations specic to a particular

kaj and to what (lat. quid). However, on the South

dialectal group.

Prior to migrations, the akavian dialects were spoken

Slavic territory, this classication is relevant only for

Croatian dialects and it results from the needs of the
Croatian linguistic community. e Slovenes use the
pronoun kaj but the Slovenian language is not a Kajkavian dialect. e Bosniaks, Montenegrins, Serbs, as
well as the Bulgarians, Macedonians and Eastern Slavs
use to, but their languages are not tokavian dialects in
the same sense as the Croatian tokavian dialect. e
Serbs, the Montenegrins and the Bosniaks do not have
this pronominal word as a criterion of dialectal classication. As far as tokavian dialects are concerned,
the archaic akavian (the so-called Slavonian) is spoken
only by Croats, Neo-tokavian ikavian and ijekavianakavian is spoken by Croats and Bosniaks, and Neotokavian ijekavian by Croats in some areas in the wider

3.3 STANDARDISATION OF
CROATIAN LANGUAGE
e millennial history of the Croatian language is attested to by texts written as early as the end of the 10th
or the beginning of the 11th century, the period in
which the three Croatian dialects (akavian, tokavian,
Kajkavian) began to form. All three Croatian dialects
played an important part in the formation of the Croatian literary language (various dialectal stylizations) and
the moulding of the Croatian linguistic culture that
led to the Croatian standard language with a tokavian
foundation.

Dubrovnik region, but also by other South Slavic peoples. Croats in Burgenland (Austria, Hungary, Slovakia)
mostly speak akavian, and rarely the tokavian or Kajkavian dialects; Croats in the Italian province of Molise
speak an archaic tokavian dialect, and Karaevo Croats

Did you know that the etymology of the

word cravatte (tie) comes from Croatian
and from French in 17th century it spread
to other languages?

in Romania speak a Torlak dialect.

Due to numerous, oen forced migrations, the areal dis-

e rst clear trends towards the shaping of the Croa-

tribution of certain Croatian dialects has changed dras-

tian standard language became apparent in the 17th

tically since the Middle Ages. Both akavian and Ka-

century, when the majority of the Croatian ethnic

jkavian were historically distributed throughout a much

community especially aer the grammar and other

53

2: Map of Croatian dialects in the Republic of Croatia

54

works of Bartol Kai (15751650) and a ourishing of Renaissance and Baroque literature from tokavian Dubrovnik recognised the linguistic structure
of the tokavian dialect (rstly with the ikavian jat
reex, but later with the jekavian reex) as the best

3.4 CHARACTERISTICS OF THE

CROATIAN LANGUAGE
3.4.1 Phonetics, phonology, morphonology

starting point for the construction of a supra-regional

Croatian literary language. Despite the choice of one

e phoneme inventory of the Croatian standard lan-

linguistic structure in the construction of their stan-

guage consists of 5 vowels (a, e, i, o, u) and 25 conso-

dard language, the Croats did not dismiss the achieve-

nants (m, v, n, l, r, j, nj, lj, p, b, f, s, z, c, t, d, , , ,

ments of the centuries-old linguistic cultures of vari-

, , d, h, k, g). e acoustic and articulatory charac-

ous dialectal stylisations within the Croatian literary

teristics of the vowels do not change depending on their

language (Kajkavian, tokavian, akavian, hybrid) that

placement (regardless whether in a short, long, accented

had marked its history within the Croatian ethnic com-

or unaccented syllable). In addition to these 5 vowels,

munity. Although the standardisation of the language

there also exist the syllabic r (crn black) and the diph-

of the Croats based upon the tokavian dialect began

thong ie, which is marked in writing as je/ije (djelo, odi-

very early, national linguistic unity was only achieved

jelo).

during the time of the Illyrian national revival (start-

e prosodic system consists of 4 accents (two long ac-

ing in 1835), when smaller groups of Croats who had

cents with a descending and ascending tone and two

until then expressed themselves in the Kajkavian id-

short accents with descending and ascending tone) and

iom also accepted the tokavian Croatian standard lan-

unaccented post-accentual lengths. e accentual sys-

guage. roughout most of the 20th century, the Croat-

tem of the Croatian standard language is neo-tokavian,

ian standard language developed in various South Slavic

although it exists today with many dierentiations from

state units under various names, and was presented

the prosodic models codied in the second half of the

as a variant of the so-called Croato-Serbian (Serbo-

19th century. Accent location is not xed to a spe-

Croatian) language. is was abandoned during the

cic syllable, but the distribution of accents does have

socio-political changes of 1990.

some limitations (e. g., the last syllable of a multi-syllable

word cannot in principle be accentuated, descending
accents are realised only in the initial syllables of noncompound words). ese rules are broken in everyday
speech, especially in large urban centres that are not lo-

Dierent stylisations of the Croatian language were

cated in Neo-tokavian regions (e. g., kontinuitt / kon-

shaped in diaspora long in the past (e. g., Burgenland

tinutt). Accent and length can have a dierentiating

Croatian, Molise Croatian). Croatian written culture is

role as they occasionally dierentiate the meaning of

marked by the use of three alphabets (Glagolitic, Cyril-

lexemes or their wordforms, e. g., gr d (= hail) : grd

lic, Latin), and the Latin script has been the foremost of

(= town, city), n (Gen. sing.) : ne (Nom. plur.).

the three among the Croats since the 16th century. Its

In Croatian some words do not have their own accent

usage was neither normed nor systematised until 1835,

(clitic), but in an accentual unit proclitics can carry an

when Ljudevit Gaj gave the Croatian Latin alphabet its

accent passed over from an accented word with a de-

modern-day form.

scending accent in the initial syllable (grd :

grd),

55

while enclitics cannot do this. e transfer of an accent

fect, pluperfect, future 1, future 2). e verbs biti (to

onto a proclitic is becoming ever more rare in everyday

be) and htjeti (to will) are auxiliary in Croatian. Verbs

speech, especially in urban centres not located in neo-

also have a complicated aspectual system (imperfective

tokavian regions.

and perfective with additional subvalues such as inchoa-

e Croatian standard language is characterised by

tivity, iterativity etc.) and they also encode the feature

a number of phonologically (Nom. sing. sladak :

of transitivity. Adjectives and adverbs can take compar-

Gen. sing. slatkoga, Nom. sing. dio : Gen. sing. di-

ative forms (three values: positive, comparative, superla-

jela) and morphonologically conditioned alternations

tive). Declension has two main types: noun declension

(Nom. sing. majka : Dat. sing. majci, Nom. sing. junak :

(nouns and indenite form of adjectives) and pronoun-

Voc. sing. junae).

adjective declension (pronouns, denite form of adjec-

Regional implementation of the Croatian standard lan-

tives, numbers). Each noun gender has its own declen-

guage is oen inuenced in speech by dialects located

sion (a-type for masculine and neuter gender, e-type for

in a given region, e. g., in the akavian Kvarner region

feminine gender), and there is a special i-type (feminine

the prevalence of the plosive t in place of the voiceless

gender nouns).

africate , or in the northwestern (Kajkavian) region, the

Suxes for noun declension are shown in Figure 3 and

non-dierentiation of and d.

for adjective-pronoun declension are shown in Figure 4.

Words in Croatian are formed by derivation and com-

3.4.2 Morphology

pounding. ere are a few dierent methods of forma-

e Croatian standard language dierentiates between

(do-iot-an), compound non-sux formation (plai-

ten parts of speech, of which ve inect (nouns, ad-

drug), compound sux formation (vanjsk-o-politiki),

jectives, numbers, pronouns, verbs) and four do not

coalescence (uz-brdo), formation through compound

inect (prepositions, conjunctions, particles, exclama-

abbreviations (Varteks) and conversion (mlada). Sux

tions), while adverbs inect only in comparation.

formation is the most common.

tion: sux formation (star-ac), prex-sux formation

Grammatical categories that characterise the majority of

declinable words are gender (three values: masculine,
neuter, feminine), number (two values: singular, plu-

3.4.3 Vocabulary, phraseology, terminology

ral), case (seven values: nominative, genitive, dative, ac-

e foundational lexical layer of the Croatian standard

cusative, vocative, locative, instrumental). Some declin-

language, aside from proto-Slavonic lexical heritage,

able words have special categories (e. g., deniteness is

consists of tokavian vocabulary with an admixture of

marked on adjectives with a full set of inectional end-

vocabulary from other Croatian dialects or vocabulary

ings; animacy is marked by ending in masculine nouns

inherited from the literary language of various dialec-

and adjectives; nouns can be concrete, material, cate-

tal stylisations from older periods (e. g., from Kajka-

gorial or collective etc.). Words that are conjugated

vian, kukac, hlae, rjenik, or akavian, spuva). Aside

(verbs) are characterised by the categories of: manner

from this, the Croatian language as a whole bears wit-

(four values: indicative, imperative, conditional, opta-

ness to direct and indirect contact with other cultures.

tive), person (three values: 1st, 2nd, 3rd), number (two

e Croatian language stands out among the remaining

values: singular, plural), voice (two values: active, pas-

South Slavic languages in signicant lexical inuence re-

sive), tense (seven values: present, aorist, imperfect, per-

ceived from Romance languages (substrate traces of the

56

Noun declension

N and G singular

N plural

a-type masculine
a-type neuter
e-type feminine
i-type feminine

opis, opisa
sunce, sunca
ena, ene
no, noi

opisi
sunca
ene
noi

3: Noun declension in the Croatian language

Case

Masculine

Neuter

Feminine

Singular
N
G
D
A
V
L
I

-i
-og(a) -eg(a)
-om(u/e) -em(u/e)
=N/=G
=N
-om(u/e) -em(u/e)
-im

-o -e
-og(a) -eg(a)
-om(u/e) -em(u/e)
=N
=N
-om(u/e) -em(u/e)
-im

-a
-e
-oj
-u
=N
=D
-om

Plural
N
G
D
A
V
L
I

-i
-ih
-im(a)
-e
=N
=D
=D

-a
-ih
-im(a)
=N
=N
=D
=D

-e
-ih
-im(a)
=N
=N
=D
=D

4: Adjective-pronoun declension in the Croatian language

57

Dalmatic language, e. g., jarbol, tunj). Italian signi-

Continuity from ancient times to the modern-day Croa-

cantly inuenced the coastal regions of Croatia (espe-

tian standard language and the participation of three di-

cially the parts formerly under Venetian control), while

alects in the construction of the Croatian standard lan-

German and, to an extent, Hungarian inuenced the

guage can be seen in its well-developed and rich phrase-

continental part.

ology (e. g., in his 16th century stylised texts, Maruli

uses the phraseme zgubiti glas = to be ashamed, to

e Church Slavonic literary language le traces in

lose face, while Zorani uses the phraseme u magnutje

older historical periods of the Croatian language, and

oka = immediately, which are nearly the same as the

so it did not present a great inuence during the time in

phrasemes izgubiti glas and u trenu oka in the tokavian-

which the standard language was being shaped. Russian

structured standard language).

did not leave as a deep mark on Croatian as it did on

Terminology in specic professional elds began to de-

the neighbouring Serbian standard language. e inu-

velop as early as the 16th century, conrmed by the

ence of the vocabulary of classical languages (Latin and

numerous Croatian (mostly multi-lingual) dictionaries

Greek) is omnipresent in Croatian culture, especially

compiled from the 16th to the 20th century. In the 19th

in intellectual vocabulary, and scientic terminology.

century, German and Czech had especially strong inu-

During the middle-Croatian period (16th to 18th cen-

ence on Croatian terminology, and English has today

tury), Turkish loan words intensively entered the Croat-

assumed this role.

ian language, especially words related to everyday life. It

is interesting to note that Burgenland Croatian, due to

3.4.4 Syntax

early migrations, does not have any Turkish loanwords,

not even those that are in standard Croatian no longer

e Croatian language belongs to a group of languages

perceived as foreign words (e. g., bubreg, izma, jastuk,

characterised by an SVO syntactic structure (Marija

etc.). In contrast to those loan-words, Burgenland Croa-

oli Ivana) and relatively free word order (numerous

tian uses older Croatian words of common Slavic origin

permutations of constituents are possible with some

and is therefore very important for the history of Croa-

limitations, such as clitic placement). As concerns the

tian lexical inventory. German and French once had

information structure of sentences, it is a basic rule

an inuence on Croatian vocabulary, and in the second

for structuring stylistically unmarked discourse that the

half of the 20th century, the inuence of English has

rst place is taken by the theme (old information), which

been ever stronger. e Czech language, although not

is followed by the rheme (comment, new information).

in direct contact, has had a strong inuence on Croat-

e subject of a sentence does not have to be explic-

ian vocabulary in several episodes, especially in the 19th

itly stated, and its omission is desirable insofar as it is

century in professional terminology enriched by Bo-

repeated a number of times within a narrow context.

goslav ulek (e. g., asopis, kisik, duik, odik). During

Double-negation is required (Nitko ga nije olio). e

the period of Yugoslavia, Croatian was inuenced by

agreement of components in gender, number and case

the Serbian language, especially because of common fed-

is typical of Croatian sentence structure.

eral state administration. Purist tendencies in vocabu-

ere are seven cases in the Croatian standard language,

lary came about occasionally from the 16th to the 20th

and case forms are combined with prepositions (obliga-

century (e. g., Zorani, Ritter Vitezovi, Reljkovi, the

tory for the locative case). An important characteristic

period from 1941 to 1945).

of Croatian verbs is their aspect while verb forms also ex-

58

press both tense and modal meaning. Sentence organisation can be both coordinated and subordinated (with
the aid of conjunctions or without them). A relatively
new occurrence in the modern language is the common
use of the Slavonic genitive (Nije olio vina), genitive expressions of possession are avoided in favour of possesive
adjectives (majina kua instead of kua majke), and the
use of preterite tenses is reduced (imperfect, aorist and
pluperfect). In modern Croatian passive constructions
are rarer than in the older Croatian language.

3.4.6 Onomastics
Croatian names represent important linguistic monuments of the linguistic, cultural and social heritage of
the people who created them. us, both personal
names (anthroponyms) and place names (toponyms) are
an important segment of Croatian linguistic culture.
e territory of present-day Croatia, roughly bound by
the river Drava in the North, the river Danube in the
East and the Adriatic Sea in the South, is very picturesquely reected in its complex stratication of geographical names. e complex stratication of Croa-

3.4.5 Orthography

tian toponymy reects centuries of coexistence of the

Although the history of Croatian culture has been

coast of the Adriatic and its hinterland throughout his-

marked by the use of three scripts (Glagolitic, Cyrillic

tory. Centuries of linguistic interpenetration and the

and Latin script), the Latin script has been the domi-

merging of various cultural traditions have le an indeli-

nant script used by Croats since the 16th century. e

ble imprint on Croatian toponymy. Furthermore, place

Croatian Latin alphabet was not fully standardized until

names attestations are frequently the oldest witnesses to

1835, when Ljudevit Gaj gave it its current-day form. It

the oldest changes in the Croatian language itself.

various ethnic groups that have settled on the Eastern

is composed of 30 characters, of which three are double

characters (d, lj, nj), and the rest are single characters,
of which ve have diacritics (, , , , ). In academic
circles, especially in the printing of texts from Croatian

Did you know that Croats were the rst Slavic

nation to bear family names since 12th century?

written heritage, the dual-characters d, lj and nj, are replaced by , and respectively. e characters q, x, y,

Since Croatian developed across religious (pre-christian

w do not exist in the Croatian alphabet originally, al-

and christian), cultural and civilisational borders, traces

though they are being used for writing foreign names.

of both East and West have been le on Croatian names.

e Croatian Latin alphabet is shown in Figure 5.

With regards to personal names, Croatians were the rst

Croatian orthography is phonological-morphonological, Slavic nation to bear family names (since the 12th censince it presents a conuence of two orthographic

tury) along the Adriatic coast due to direct Romance

principles: dominant phonological (e. g., the mark-

cultural inuence. e oldest layer of Croatian names

ing of assimilation) and subordinate morphonological

is founded upon proto-Slavic name forms that are fol-

(e. g., podcrtati). Interword separation is logical, and not

lowing common Indo-European name formation pat-

grammatical (as it once was). It is typical of Croatian

terns. e patronymics form the basis for the largest

orthography that the writing of foreign names is not

part of inventory of family names but, unlike in Rus-

adjusted to their pronunciation or the graphic inven-

sian, they are not productive anymore and remain un-

tory of the Croatian alphabet (e. g., John, not Don, or

changed as frozen family names that are incorporated in

Washington, not Vaington).

inectional system as nouns. In contrast to the Croatian

59

Capital letters
A
L

B
LJ

C
M

NJ

D
O

D
P

E
S

G
T

H
U

I
V

J
Z

g
t

h
u

i
v

j
z

Lowercase letters
a
l

b
lj

c
m

nj

d
o

d
p

e
s

5: The Croatian Latin alphabet

toponomastic system, where we found almost no Turk-

of the common name (Serbo-Croato-Sloenian during

ish inuence, many Croatian family names were formed

the Kingdom of Yugoslavia; Serbo-Croatian or Croato-

upon Turkish loan-words with Croatian suxes, since

Serbian, Croatian or Serbian during the communist Yu-

most family names in Croatia were created aer the

goslavia). During the Second World War and a few years

Council of Trent in the 16th century, at the time when a

later all ocial documents in Yugoslavia were published

large portion of Croatian lands was under Turkish rule.

in four ocial languages (Croatian, Macedonian, Serbian, Slovenian), but soon a lot of political eort was

3.5 THE CROATIAN STANDARD

LANGUAGE AND OTHER
TOKAVIAN-STRUCTURED
LANGUAGES
e four national languages, Croatian, Serbian, and recently Bosnian and Montenegrin, all share tokavian
as structural basis, however the traditions and superstructures of these languages are fairly dierent. What
is specic to Croatians linguistic history and culture
among other South Slavic languages is the relationship between its three dialects (Kajkavian, akavian,
tokavian), which continually enriches the tokavianstructured Croatian standard language.

Because of

put again into convergence of Croatian and Serbian.

Despite all attempts to recognise the ocial existence
of Croatian as a language on its own, the forcing of unied terminology, vocabulary, orthography and other
linguistic norms in Yugoslavia, led to the ocial recognition of one standard language (Serbo-Croatian) with
two variants (eastern or Serbian and western or Croatian). e reaction from Croatia came in the form of
Declaration on the Position of the Croatian Language
that openly advocated the recognition of the independent Croatian language and was unanimously signed
in 1967 by leading scientic, cultural and educational
institutions as well as leading intellectuals throughout
Croatia who took a great risk with such an open political move in communist times.

dierent starting points (the non-existence of a basic,

common standard) and traditions in language cultiva-

In the past 20 years, the four tokavian-structured stan-

tion and standardisation, the disunity of neo-tokavian

dard languages have developed autonomously as na-

structure and dierences in linguistic superstructure,

tional standard languages in a naturally diverging way,

one monolithic standard language was never formed

and no agreement or coordination exists concerning

during the existence of the Yugoslavian states, although

their norming, which has increased dierences between

there were serveral attempts of political imposition

them.

60

3.6 LINGUISTIC CULTIVATION

IN CROATIA

proper language usage and linguistic expertise are per-

e Croatian Language Council was founded by a de-

the most frequently asked questions are available on the

cision of the Ministry of Science, Education and Sport

Language Advice Portal [10] on the Institute web site.

manent duties of the Institute. Advice is given by phone,

e-mail and in written form. Furthermore, the answers to

taken on 14th April 2005. Its basic task is the systematic

and scientic care of the Croatian standard language.
e specic tasks of the Council are:
to tend to the Croatian standard language;

The basic task of the Croatian Language Council

is the systematic and scientic care of the
Croatian standard language.

to discuss current dilemmas and open issues in the

Croatian standard language;

e Institutes STRUNA project [11], which develops

to warn of cases of infractions of the constitutional

the Croatian professional terminology, deserves a spe-

decree on the position of Croatian as the ocial lan-

cial mention. e goal of this project is to establish a

guage of the Republic of Croatia;

system of coordinating terminological work in all pro-

to promote the culture of the Croatian standard language in written and oral communication;
to tend to the status and role of the Croatian standard language in light of Croatias integration into
the European Union;
to make decisions on further standardisation processes of the Croatian standard language;
to take care of language issues and set principles for

fessional elds in Croatia, and in doing so contribute

to the improvement of the quality and eectiveness of
higher education and scientic research work through
the creation of unied and veried terminology that can
be used by experts in all elds, as well as by interested
participants from the general public. e establishment
of a research terminology network and scientic cooperation between institutions that deal in various aspects
of terminological work is also planned.

the orthographic standardization.

e Croatian Language Council meets regularly and
draws conclusions aer thorough debate. e Institute of Croatian Language and Linguistics hosts the

Today English loan words are common

in the informal language but much less so in
the formal or written language.

Council, provides technical and administrative support

as well as linguistic expertise when necessary.

Besides this, other Croatian scientic institutions (sev-

e Institute of Croatian Language and Linguistics [9]

eral universities with their departments of Croatian lan-

is the central Croatian institution for the research of

guage and literature) and cultural institutions (such

the Croatian language, and one department of the In-

as Matica hrvatska) also take part in the care of the

stitute (the Croatian Standard Language Department)

Croatian language. Public media, such as state radio-

is dedicated to the description of the Croatian standard

television and some newspaper publishers, have well-

language, with special attention paid to linguistic cul-

developed proofreading services for the Croatian stan-

ture (e. g., work on oering linguistic advice to the pub-

dard language and pay special attention to the quality

lic and the writing of language handbooks). Advice on

of language they use in their public text production.

61

3.7 LANGUAGE IN EDUCATION

is open to general public), it is also possible to study

Croatian is ocial in all primary and secondary schools,

kindergarten level to secondary education, is available

except in regions with national minority residents.

and nanced by the Croatian government for the Ser-

However, it is not dened as obligatory for the use at

bian, Czech, Hungarian and Italian minority.

Hebrew. Education on minority languages, from the

universities. ere is a pronounced tendency in Croatia, especially in so-called hard sciences to teach in the
English language. ere were agreeable opinions that it
could be functional and useful, but also harmful and unacceptable not to teach in the Croatian language at universities. It would have devastating eects for the development of the Croatian scientic terminology and occupational phraseology. erefore e Croatian Language Council advised the Ministry to legally dene
the language usage at higher education.
In primary and secondary schools, Croatian language
and literature is taught as a subject, and takes up considerable space in the curriculum. As part of this subject,
Croatian grammar, vocabulary and literature is studied,
and written and verbal expression in Croatian is devel-

3.8 INTERNATIONAL ASPECTS

e use of the Croatian standard language in countries
in the region is regulated by the laws of these countries. e status of the Croatian standard language as
one of the ocial languages of neighbouring Bosnia and
Herzegovina is especially important, and so Croatian institutions pay special attention to cooperation with scientic and cultural institutions of the Croatian nation
in Bosnia and Herzegovina. e Republic of Croatias
cultural institutions establish cooperation with many
Croatian diasporic institutions throughout the world.

oped. e PISA test, which tests the skills of pupils

at the global level, has been executed in Croatia since
2006, and the rst results of testing showed that Croatian 15-year-olds took 26th place of world countries,
placed ahead of ten European Union member states and

When Croatia joins the European Union in

2013, the Croatian language will become the
24th ocial language of the EU.

the United States of America.

Besides Croatian, in primary and secondary education
it is obligatory to study at least one foreign language

Lectures of Croatian language are organised in schools

from the fourth grade. However, English (only rarely

abroad for the children of Croatian citizens who reside

French or German) is oen taught already in kinder-

either temporarily or permanently in other countries.

gartens. English is usually the rst foreign language in

e Croatian language is taught at many foreign insti-

primary education. e most widespread second for-

tutions and Slavic studies centres (there are 36 ocial

eign language is German, then Italian and French. In

exchange instructorships for the Croatian language and

secondary education Russian and Spanish are occasion-

literature as well as 2 centres for Croatian studies in Aus-

ally taught as second or third foreign languages. Latin

tralia and Canada in the jurisdiction of and nanced by

and Old Greek are taught in all classics-program schools

the Ministry of Science, Education and Sport of the Re-

that start from the h grade of primary school. Fur-

public of Croatia). A number of centres for the study

thermore, Latin is still obligatory in all humanistic sec-

of Croatian as a second or foreign language operate in

ondary schools. In a Jewish minority school (which

Croatia, the best-known of which is Croaticum [12].

62

3.9 CROATIAN ON THE

INTERNET

comprehensive list of mono- and multilingual dictio-

According to the statistical information of the Croatian

maintained [13]. At the same Faculty a portal on Croa-

Bureau of Statistics, the use of information and commu-

tian Language Technologies [14] is maintained since

nications technology in enterprises and households are

1999.

naries, grammars and orthographies. At the Faculty of

Humanities and Social Sciences a similar web page is

shown in Figures 6 and 7.

e most-visited Croatian websites are: net.hr (a news,
sports, entertainment and events portal), index.hr (general web portal, info, services, news, sports, entertain-

The growing importance of the Internet is

important for Language Technologies.

ment, automotive, gastronomy), jutarnji.hr (the website

of the daily newspaper Jutarnji list), 24sata.hr (website
of the daily newspaper 24 sata), tportal.hr (newsportal

e Croatian-language Wikipedia was founded in 2003

of HT, Croatian Telecomm), njuskalo.hr (Njukalo

and has 108,528 articles (as of 2012-05-24), being the

advertisments portal), vecernji.hr (website of the daily

30th Wikipedia by number of ocial articles.

newspaper Veernji list), forum.hr (the largest Croa-

Access to resources in Croatian has been made easier in

tian web forum, discussing society, culture, entertain-

recent years by Croatian institutions and organisations

ment, etc.). Seven daily Croatian newspapers publish

undergoing the digitisation process (including signi-

their articles on their own dedicated portals in addition

cant projects supported by Ministry of Science, Educa-

to their paper versions.

tion and Sports and Ministy of Culture for digitising

e Institute of Croatian Language and Linguistics

Croatian cultural heritage) which has increased the vis-

maintains the web page about Croatian that features a

ibility of the Croatian language among internet sources.

63

Usage of information and communication technologies (ICT) in enterprises (%)

Computer usage
Internet access
Web site
Usage of nancial and banking services
E-government usage

2008

2009

2010

98
97
64
84
56

98
95
57
84
61

97
95
61
85
63

6: ICT in enterprises

Households equipped with information and communication technologies (ICT) (%)

Personal computer
Internet access
Mobile phone

2008

2009

2010

53
45
81

55
50
82

60
57

7: ICT in households

64

4
LANGUAGE TECHNOLOGY SUPPORT FOR
CROATIAN
Language technologies are used to develop soware

computer-assisted language learning

systems designed to handle human language and are

information retrieval

therefore oen called human language technologies.

information extraction

Human language comes in spoken and written forms.

text summarisation

While speech is the oldest and in terms of human evolution the most natural form of language communication, complex information and most human knowledge
is stored and transmitted through the written word.

question answering
speech recognition
speech synthesis

Speech and text technologies process or produce these

Language technology is an established area of research

dierent forms of language, using dictionaries, rules

with an extensive set of introductory literature. e in-

of grammar, and semantics. is means that language

terested reader is referred to the following references:

technology (LT) links language to various forms of

[16, 17, 18, 19].

knowledge, independently of the media (speech or text)

Before discussing the above application areas, we will

in which it is expressed. Figure 1 illustrates the LT land-

briey describe the architecture of a typical LT system.

scape.
When we communicate, we combine language with
other modes of information and communication media
for example speaking can involve gestures and facial
expressions. Digital texts link to pictures and sounds.
Movies may contain language in spoken and written
form. In other words, speech and text technologies overlap and interact with other multimodal communication
and multimedia technologies.
In this section, we will discuss the main application
areas of language technology, i. e., language checking,
web search, speech interaction, and machine translation. ese applications and basic technologies include:

4.1 APPLICATION
ARCHITECTURES
Soware applications for language processing typically
consist of several components that mirror dierent aspects of language. While such applications tend to be
complex, Figure 2 shows a highly simplied architecture
of a typical text processing system. e rst three modules handle the structure and meaning of the text input:
1. Pre-processing: cleans the data, analyses or removes
formatting, detects the input languages, and so on.
2. Grammatical analysis: nds the verb, its objects,

spelling correction

modiers and other sentence elements; detects the

authoring support

sentence structure.

65

Speech Technologies
Multimedia &
Multimodality
Technologies

Language
Technologies

Knowledge Technologies

Text Technologies

8: Language technologies

3. Semantic analysis: performs disambiguation (i. e.,

LT for the Croatian language is summarised in Figure 14

computes the appropriate meaning of words in a

(p. 78) at the end of this chapter. is table lists all tools

given context); resolves anaphora (i. e., which pro-

and resources that are boldfaced in the text. LT support

nouns refer to which nouns in the sentence); rep-

for Croatian is also compared to other languages that are

resents the meaning of the sentence in a machine-

part of this series.

readable way.
Aer analysing the text, task-specic modules can per-

4.2 CORE APPLICATION AREAS

form other operations, such as automatic summarisa-

In this section, we focus on the most important LT tools

tion and database look-ups.

and resources, and provide an overview of LT activities

In the remainder of this section, we rstly introduce

in Croatia.

the core application areas for language technology, and

follow this with a brief overview of the state of LT re-

4.2.1 Language Checking

search and education today, and a description of past

Anyone who has used a word processor such as Mi-

and present research programmes. Finally, we present

croso Word knows that it has a spell checker that high-

an expert estimate of core LT tools and resources for

lights spelling mistakes and proposes corrections. e

Croatian in terms of various dimensions such as avail-

rst spelling correction programs compared a list of ex-

ability, maturity and quality. e general situation of

tracted words against a dictionary of correctly spelled

Input Text

Pre-processing

Output

Grammatical Analysis

Semantic Analysis

Task-specific Modules

9: A typical text processing architecture

66

Statistical Language Models

Input Text

Spelling Check

Grammar Check

Correction Proposals

10: Language checking (top:statistical; bottom:rule-based)

words. Today these programs are far more sophisticated.

around data from English. Neither approach can trans-

Using language-dependent algorithms for grammatical

fer easily to Croatian because the language has a exi-

analysis, they detect errors related to morphology (e. g.,

ble word order and rich inection that contribute abun-

plural formation) as well as syntax-related errors, such

dantly to the data sparsness problem in such systems.

as a missing verb or a conict of verb-subject agreement

Language Checking is not limited to word processors;

(e. g., she *write a letter). However, most spell checkers

it is also used in authoring support systems, i. e., so-

will not nd any errors in the following text [66]:

ware environments in which manuals and other types

I have a spelling checker,

It came with my PC.
It plane lee marks four my revue
Miss steaks aye can knot sea.

of technical documentation for complex IT, healthcare,

engineering and other products, are written. To oset customer complaints about incorrect use and damage claims resulting from poorly understood instructions, companies are increasingly focusing on the qual-

Handling these kinds of errors usually requires an analy-

ity of technical documentation while targeting the in-

sis of the context. For example: deciding if an Croatian

ternational market (via translation or localisation) at

noun should be written with capital rst letter (female

the same time. Advances in natural language process-

personal name) or not (common noun), as in:

ing have led to the development of authoring support

soware, which helps the writer of technical documen-

Slatka je ova vinja. [is cherry is sweet.]

tation to use vocabulary and sentence structures that are

Slatka je ova Vinja. [is Cherry is sweet.]

consistent with industry rules and (corporate) terminol-

is type of analysis either needs to draw on languagespecic grammars laboriously coded into the soware

ogy restrictions, but such systems are not yet available

for Croatian.

by experts, or on a statistical language model. In this

case, a model calculates the probability of a particular
word as it occurs in a specic position (e. g., between

Language checking is not limited to word

processors but also applies to authoring systems.

the words that precede and follow it). For example: jaz
izmeu (gap between) is much more probable word se-

Although the research on computational models

quence than jaz generacija (generation gap). A statisti-

of inectional morphology existed in 1980s the

cal language model can be automatically created by us-

rst industry-strength spelling checker for Croatian

ing a large amount of (correct) language data, a text cor-

Hrvatski raunalni pravopis has been published in 1996

pus. Most of these two approaches have been developed

[8]. Soon aer it was bought by Microso and today

67

it represents the integral part of Croatian MS Oce

the wordforms in which Croatian lexemes could appear

proong tools and it is widely used. Other spelling

in texts. Unlike the, e. g., English nouns where only four

checkers have also been developed by several private

wordforms are possible for a noun lexeme (hand, hands,

companies, but none of them has been so success-

hands, hands) in Croatian theoretically it can appear

ful. An on-line Croatian Academic Spelling Checker

in 14 dierent wordforms, but they are represented on

(Hascheck) [21] exists since 1994 and is still in use. An

average with 10 dierent types (ruka, ruke, ruci, ruku,

open source spelling checker for Croatian also exists,

rukom, rukama ). Google can retrieve forms like ruka,

it can be used with OpenOce on dierent operating

ruke, but ruci is still not connected to the lemma ruka.

systems and is based on Ispell/Aspell. ese programs

ere is room for improvement when Google has to deal

are based on the very large lexicon of correct wordforms

with inectionally rich languages where lexemes appear

which have two drawbacks: 1) strings that represent

in many dierent wordforms. e Google success story

correct wordforms appearing in a wrong co text; 2) the

shows that a large volume of data and ecient indexing

inability to distinguish between real spelling errors and

techniques can deliver satisfactory results using a mainly

wordforms which are correct, but which are unknown

statistical approach to language processing, but they also

to the lexicon. Besides spell checkers and authoring sup-

depend heavily on the language structure.

port, Language Checking is also important in the eld

of computer-assisted language learning and is applied
to automatically correct queries sent to Web Search engines, e. g., Googles Did you mean suggestions.

4.2.2 Web Search

The next generation of search engines

will have to include much more sophisticated
language technology.

For more sophisticated information requests, it is essen-

Searching the Web, intranets or digital libraries is proba-

tial to integrate deeper linguistic knowledge to facilitate

bly the most widely used yet largely underdeveloped lan-

text interpretation. Experiments using lexical resources

guage technology application today. e Google search

such as machine-readable thesauri or ontological lan-

engine, which started in 1998, now handles about

guage resources (e. g., WordNet for English or Croatian

80% of all search queries [22]. Since 2004, the verb

Wordnet, CroWN for Croatian) have demonstrated im-

guglati/googlati and its derivatives (iz-/na-/pre-/pro-/u-

provements in nding pages using synonyms of the orig-

)guglati/(iz-/na-/pre-/pro-/u-)googlati is used in Croat-

inal search terms, such as nuklearna energija and atom-

ian, even though it has not made its way into printed

ska energija (nuclear energy and atomic energy) or even

dictionaries (even more complex derivatives such as

more loosely related terms.

ugugljiv googlable are recorded). e Google search

e next generation of search engines will have to in-

interface and results page display has not signicantly

clude much more sophisticated language technology,

changed since the rst version. However, in the current

especially to deal with search queries consisting of a

version, Google oers spelling correction for misspelled

question or other sentence type rather than a list of key-

words and incorporates basic semantic search capabili-

words. For the query, Give me a list of all companies

ties that can improve search accuracy by analysing the

that were taken over by other companies in the last ve

meaning of terms in a search query context [23]. With

years, a syntactic as well as semantic analysis is required.

the help of this algorithm it also started to cover some of

e system also needs to provide an index to quickly re-

68

Web Pages

Pre-processing

Semantic Processing

Indexing
Matching
&
Relevance

Pre-processing

Query Analysis

User Query

Search Results

11: Web search

trieve relevant documents. A satisfactory answer will re-

Now that data is increasingly found in non-textual for-

quire syntactic parsing to analyse the grammatical struc-

mats, there is a need for services that deliver multime-

ture of the sentence and determine that the user wants

dia information retrieval by searching images, audio les

companies that have been acquired, rather than compa-

and video data. In the case of audio and video les,

nies that have acquired other companies. For the expres-

a speech recognition module must convert the speech

sion last ve years, the system needs to determine the

content into text (or into a phonetic representation)

relevant range of years, taking into account the present

that can then be matched against a user query.

year. e query then needs to be matched against a huge

amount of unstructured data to nd the pieces of information that are relevant to the users request. is process is called information retrieval, and involves searching and ranking relevant documents. To generate a list
of companies, the system also needs to recognise a particular string of words in a document represents a company name, using a process called named entity recognition. A more demanding challenge is matching a query
in one language with documents in another language.
Cross-lingual information retrieval involves automatically translating the query into all possible source languages and then translating the results back into the
users target language.

For inectional languages like Croatian it is important

to be able to search for all the inectional forms of a
word at once, instead of having to enter each dierent form separately. is can be done with the aid of
the Croatian Lemmatisation Server that has been developed at the Department of Linguistics, Faculty of Humanities and Social Sciences at the University of Zagreb and is freely accessible [24] providing an interface
to the Croatian Morphological Lexicon, a comprehensive full wordforms database. It contains over 110,000
lexemes yielding over 4 million inectional wordforms
where each entry contains lemma, wordform and full
MSD tag and it is MULTEXT East [25] compliant.

69

In 2009 as a result of a joint Flemish-Croatian project

One of the major challenges of ASR systems is to ac-

CADIAL [26], the governmental agency HIDRA en-

curately recognise the words a user utters. is means

abled the public web access to all Croatian legislative

restricting the range of possible user utterances to a

documents using the inectionally sensitive search en-

limited set of keywords, or manually creating language

gine [27]. is engine also enables cross-lingual docu-

models that cover a large range of natural language ut-

ment retrieval since all documents are indexed with EU-

terances. Using machine learning techniques, language

ROVOC descriptors thus allowing the usage of English

models can also be generated automatically from speech

EUROVOC descriptors in queries.

corpora, i. e., large collections of speech audio les and

text transcriptions. Restricting utterances usually forces

4.2.3 Speech Interaction

Speech interaction is one of many application areas that
depend on speech technology, i. e., technologies for processing spoken language. Speech interaction technology is used to create interfaces that enable users to interact in spoken language instead of using a graphical
display, keyboard and mouse. Today, these voice user
interfaces (VUI) are used for partially or fully automated telephone services provided by companies to customers, employees or partners. Business domains that
rely heavily on VUIs include banking, supply chain,
public transportation, and telecommunications. Other
uses of speech interaction technology include interfaces
to car navigation systems and the use of spoken language
as an alternative to the graphical or touchscreen interfaces in smartphones.
Speech interaction technology comprises four tech-

people to use the voice user interface in a rigid way and

can damage user acceptance; but the creation, tuning
and maintenance of rich language models will signicantly increase costs. VUIs that employ language models and initially allow a user to express their intent more
exibly prompted by a How may I help you? greeting
tend to be automated and are better accepted.
Companies tend to use utterances pre-recorded by professional speakers for generating the output of the voice
user interface. For static utterances where the wording does not depend on particular contexts of use or
personal user data, this can deliver a rich user experience. But more dynamic content in an utterance may
suer from unnatural intonation because dierent parts
of audio les have simply been strung together. rough
optimisation, todays TTS systems are getting better at
producing natural-sounding dynamic utterances.

nologies:
1. Automatic speech recognition (ASR) determines
which words are actually spoken in a given sequence
of sounds uttered by a user.
2. Natural language understanding analyses the syntactic structure of a users utterance and interprets it according to the system in question.
3. Dialogue management determines which action to
take given the user input and system functionality.

Speech interaction is the basis for creating

interfaces that allow a user to interact with
spoken language instead of a graphical
display, keyboard and mouse.
Interfaces in speech interaction have been considerably
standardised during the last decade in terms of their various technological components. ere has also been
strong market consolidation in speech recognition and
speech synthesis. e national markets in the G20 coun-

4. Speech synthesis (text-to-speech or TTS) trans-

tries (economically resilient countries with high popu-

forms the systems reply into sounds for the user.

lations) have been dominated by just ve global play-

70

Speech Output

Speech Input

Speech Synthesis

Signal Processing

Phonetic Lookup &

Intonation Planning

Natural Language
Understanding &
Dialogue

Recognition

12: Speech-based dialogue system

ers, with Nuance (USA) and Loquendo (Italy) being the

most prominent players in Europe. In 2011, Nuance announced the acquisition of Loquendo, which represents
a further step in market consolidation.

4.2.4 Machine Translation

e idea of using digital computers to translate natural
languages can be traced back to 1946 and was followed
by substantial funding for research during the 1950s and

Although the Croatian diphone base was developed

again in the 1980s. Yet machine translation (MT) still

within the MBROLA [28] project in 1998 in which

cannot meet its initial promise of across-the-board au-

the Department of Phonetics, Faculty of Humanities

tomated translation.

and Social Sciences, University of Zagreb participated,

up to now, there has been no commercial application
of Croatian TTS or ATS systems developed in Croatia. Research in this eld has been conducted also at
the Faculty of Electrical Engineering and Computing of

At its basic level, Machine Translation

simply substitutes words in one natural language
with words in another language.

the same university [29] as well as at the University of

Rijeka where a strong group works on the development
of resources and tools for speech processing of Croatian
[3, 31, 32].

e most basic approach to machine translation is the

automatic replacement of the words in a text written
in one natural language with the equivalent words of

Looking ahead, there will be signicant changes, due to

another language. is can be useful in subject do-

the spread of smartphones as a new platform for man-

mains that have a very restricted, formulaic language

aging customer relationships, in addition to xed tele-

such as weather reports. However, in order to produce a

phones, the Internet and e-mail. is will also aect

good translation of less restricted texts, larger text units

how speech interaction technology is used. In the long

(phrases, sentences, or even whole passages) need to be

term, there will be fewer telephone-based VUIs, and

matched to their closest counterparts in the target lan-

spoken language apps will play a far more central role

guage. e major diculty is that human language is

as a user-friendly input for smartphones. is will be

ambiguous. Ambiguity creates challenges on multiple

largely driven by stepwise improvements in the accu-

levels, such as word sense disambiguation at the lexical

racy of speaker-independent speech recognition via the

level (a jaguar is a brand of car or an animal) or the as-

speech dictation services already oered as centralised

signment of the prepositional phrases on the syntactic

services to smartphone users.

level, e. g., as in:

71

Source Text

Text Analysis (Formatting,

Morphology, Syntax, etc.)

Statistical
Machine
Translation

Translation Rules
Target Text

Text Generation

13: Machine translation (left:statistical; right:rule-based)

Policajac je uoio ovjeka bez teleskopa.

[e policeman spotted a man without a telescope.]
Policajac je uoio ovjeka bez pitolja.
[e policeman spotted a man without a pistol.]
One way to build an MT system is to use linguistic rules. For translations between closely related languages, a translation using direct substitution may be
feasible in cases such as the above example. However,
rule-based (or linguistic knowledge-driven) systems often analyse the input text and create an intermediary
symbolic representation from which the target language

like knowledge-driven systems, however, statistical (or

data-driven) MT systems oen generate ungrammatical output. Data-driven MT is advantageous because
less human eort is required, and it can also cover special particularities of the language (e. g., idiomatic expressions) that are oen ignored in knowledge-driven
systems. Regarding the European languages, acceptable
translations can be obtained for English and the Romance languages, but the quality is downgraded substantially for Germanic, Slavic, Finno-Ugric and Baltic
languages [34].

text can be generated. e success of these methods is

highly dependent on the availability of extensive lexicons with morphological, syntactic, and semantic information, and large sets of grammar rules carefully designed by skilled linguists. is is a very long and there-

Machine Translation is particularly

challenging for Slavic languages because of
their free word order, inectional richness and
long distance dependencies.

fore costly process.

In the late 1980s when computational power increased

e strengths and weaknesses of knowledge-driven and

and became cheaper, interest in statistical models for

data-driven machine translation tend to be complemen-

machine translation began to grow. Statistical models

tary, so that nowadays researchers focus on hybrid ap-

are derived from analysing bilingual text corpora, paral-

proaches that combine both methodologies. One such

lel corpora, such as the Europarl parallel corpus, which

approach uses both knowledge-driven and data-driven

contains the proceedings of the European Parliament in

systems, together with a selection module that decides

21 European languages or JRC Acquis parallel corpus in

on the best output for each sentence. However, results

22 European languages [67]. Given enough data, statis-

for sentences longer than, say, 12 words, will oen be

tical MT works well enough to derive an approximate

far from perfect. A more eective solution is to com-

meaning of a foreign language text by processing paral-

bine the best parts of each sentence from multiple out-

lel versions and nding plausible patterns of words. Un-

puts; this can be fairly complex, as corresponding parts

72

of multiple alternatives are not always obvious and need

nian and Romanian), and propose novel approaches for

to be aligned.

adapting existing MT technologies to specic narrow

For Croatian, MT is particularly challenging. e free

domains, signicantly increasing language and domain

word order and extensive inection is a challenge for

coverage of automated translation.

generating words with proper endings that mark gram-

e LetsMT! project [41] builds an innovative online

matical categories of gender, case, number, mood, tense,

collaborative platform for data sharing and MT genera-

etc. Also the required agreement in all these categories

tion. is cloud-based platform provides all categories

between e. g., attributes and their nouns or only in num-

of users with an opportunity to upload their proprietary

ber and gender for subject and predicate represent addi-

resources to the repository and receive a tailored statis-

tional challenge.

tical MT system trained on such resources. e latter

Although the pioneering workshop on machine translation was organised at the University of Zagreb, Faculty of Humanities and Social Sciences by eljko Bujas
and Bulcs Lszl as early as 1959 [35], no serious research on MT for Croatian happened until the beginning of 21st century. e nationally funded project Information Technology in Translation and e-Learning

can be shared with other users who can exploit them further on. e translation services of the LetsMT! project
can be used in several ways: through the web portal,
through a widget provided for free inclusion in a webpage, through browser plug-ins, and through integration in computer-assisted translation (CAT) tools and
dierent online and oine applications.

of Croatian [36] started in 2007 with the goal to in-

Google Translate has oered translations to and from

vestigate the prerequisites in building MT systems for

Croatian since 2008. e quality of the translations was

translation into and from Croatian. Starting in 2010

rather poor in the beginning, but is getting better as

several EC co-funded projects were undertaken to ad-

more and more parallel Croatian-English data is avail-

vance research and development of machine transla-

able on-line.

tion for under-resourced languages, including Croatian.

ere is still a huge potential for improving the qual-

ese projects ICT-PSP project LetsMT! [37] and

ity of MT systems. e challenges involve adapting lan-

FP7 project ACCURAT [38] are developing inno-

guage resources to a given subject domain or user area,

vative methods for making it easier to gather data for

and integrating the technology into workows that al-

MT and to create customized MT systems for dier-

ready have term bases and translation memories. An-

ent domains and usage scenarios. In both projects the

other problem is that most of the current systems are

group from the Faculty of Humanities and Social Sci-

English-centred and only support a few languages from

ences, University of Zagreb is taking part.

and into Croatian. is leads to friction in the trans-

e ACCURAT project [39] researches novel methods

lation workow and forces MT users to learn dierent

that exploit comparable corpora to compensate for the

lexicon coding tools for dierent systems.

shortage of linguistic resources to improve MT qual-

Evaluation campaigns help to compare the quality of

ity for under-resourced languages and narrow domains

MT systems, the dierent approaches and the status

[40]. e ACCURAT projects target is to achieve

of the systems for dierent language pairs. Figure 14

strong improvement in translation quality for a number

(p. 33), which was prepared during the EC Euromatrix+

of new EU ocial languages and languages of associated

project, shows the pair-wise performances obtained for

countries (Croatian, Estonian, Greek, Latvian, Lithua-

22 of the 23 ocial EU languages (Irish was not com-

73

pared). e results are ranked according to a BLEU

analysed and compared (do they provide conicting an-

score, which indicates higher scores for better transla-

swers?); and how specic information (the answer) can

tions [33]. A human translator would normally achieve

be reliably extracted from a document without ignoring

a score of around 80 points.

the context.

e best results (shown in green and blue) were achieved

by languages which benet from considerable research
eorts, within coordinated programs, and from the existence of many parallel corpora (e. g., English, French,
Dutch, Spanish, German), the worst (in red) by lan-

Language technology applications often

provide signicant service functionalities behind
the scenes of larger software systems.

guages that are very dierent from other languages (e. g.,
Hungarian, Maltese, Finnish).

uestion answering is in turn related to information

extraction (IE), an area that was extremely popular and

4.2.5 Other application areas

inuential when computational linguistics took a statis-

Building language technology applications involves a

cic pieces of information in specic classes of docu-

range of subtasks that do not always surface at the level

ments, such as the key players in company takeovers as

of interaction with the user, but they provide signicant

reported in newspaper stories. Another common sce-

service functionalities behind the scenes of the system

nario that has been studied is reports on terrorist inci-

in question. ey all form important research issues

dents. e task here consists of mapping appropriate

that have now evolved into individual sub-disciplines of

parts of the text to a template that species the perpe-

computational linguistics. uestion answering, for ex-

trator, target, time, location and results of the incident.

ample, is an active area of research for which annotated

Domain-specic template-lling is the central charac-

corpora have been built and scientic competitions have

teristic of IE, which makes it another example of a be-

been initiated. e concept of question answering goes

hind the scenes technology that forms a welldemar-

beyond keyword-based searches (in which the search en-

cated research area, which in practice needs to be em-

gine responds by delivering a collection of potentially

bedded into a suitable application environment.

relevant documents) and enables users to ask a concrete

In 2009 the Croatian Newswire Agency (HINA) [42]

question to which the system provides a single answer.

started to develop a system for (pre)processing of their

For example:

news streams that included lemmatisation, named en-

tical turn in the early 1990s. IE aims to identify spe-

tity recognition [68] and classication, classication of

Question: How old was Neil Armstrong when he

news to a predened topic schema and keyword extrac-

stepped on the moon?

tion. is system was developed jointly by the Faculty

Answer: 38.

of Electrical Engineering and Computing [43] and the

Faculty of Humanities and Social Sciences, both from

While question answering is obviously related to the

the University of Zagreb.

core area of web search, it is nowadays an umbrella term

Text summarisation and text generation are two bor-

for such research issues as which dierent types of ques-

derline areas that can act either as standalone applica-

tions exist, and how they should be handled; how a set

tions or play a supporting role. Summarisation attempts

of documents that potentially contain the answer can be

to give the essentials of a long text in a short form, and

74

is one of the features available in Microso Word. It

in the Croatian higher education system as an inde-

mostly uses a statistical approach to identify the im-

pendent subject of studying. However, at the Depart-

portant words in a text (i. e., words that occur very fre-

ment of Linguistics, Faculty of Humanities and Social

quently in the text in question but less frequently in gen-

Sciences the Algebraic linguistic approaches have been

eral language use) and determine which sentences con-

studied continuously since 1950s, and it was during the

tain the most of these important words. ese sen-

Bologna reform in higher education in 2005 that the

tences are then extracted and put together to create the

Language Technologies topics were collected together

summary. In this very common commercial scenario,

in the special study direction of Computational Lin-

summarisation is simply a form of sentence extraction,

guistics at the two-year Masters programme in Linguis-

and the text is reduced to a subset of its sentences. An

tics at the same department. A similar programme was

alternative approach, for which some research has been

launched at the University of Zadar in 2010.

carried out, is to generate brand new sentences that do

not exist in the source text.

For the Croatian language, research in most text

technologies is much less developed than for
other European languages.

4.4 NATIONAL PROJECTS AND

INITIATIVES
ere are only about 5.5 million people speaking Croatian, and this is not enough to sustain costly development
of new commercial products. It costs just as much to

is requires a deeper understanding of the text, which

means that so far this approach is far less robust. On the
whole, a text generator is rarely used as a stand-alone application but is embedded into a larger soware environment, such as a clinical information system that collects,
stores and processes patient data. Creating reports is just
one of many applications for text summarisation. None
of these technologies exist for Croatian apart from iso-

build language resources and tools for Croatian as for

languages with hundreds of millions of speakers. As a
result, the number of commercial companies in the language technology industry in Croatian is close to zero.
e role of the main funder of language technology research was partially taken by the state, but certainly not
to the extent necessary to develop all the resources and
tools needed.

lated experiments that have been performed on text

summarisation [44] and generation [45].

4.3 EDUCATIONAL
PROGRAMMES

Did you know that the rst usage of a computer

parallel corpus in contrastive linguistic in the
history of linguistics was done in Zagreb in 1968?

In Croatia activities for collecting language resources,

Language technology is a very interdisciplinary eld

i. e., computer corpora, started as early as 1967 when

that involves the combined expertise of linguists, com-

the rst computer corpus of Croatian text was collected

puter scientists, mathematicians, philosophers, psy-

by eljko Bujas and its concordance produced [46] at

cholinguists, and neuroscientists among others. As a re-

the Institute of Linguistics, Faculty of Humanities and

sult, it has not acquired a clear, independent existence

Social Sciences of the University of Zagreb. Since then,

75

this institution has become a central institution for cor-

puting, most research activities in the elds of computa-

pus linguistics research in Croatia. In 1968 the rst us-

tional linguistics, corpus linguistics and language tech-

age of computer parallel corpus in contrastive linguis-

nology today are funded by the Ministry of Science, Ed-

tics ever, was led by Rudolf Filipovi [47]. e com-

ucation and Sports through LT related projects. e

puter processing of old Croatian authors was going on in

rst one Computational Processing of the Croatian Liter-

1970s and 1980s while the collection of the One-million

ary Language started in 1991, and was followed in 1996

corpus of Croatian literary language started in 1976, lead

by Computational Processing of the Croatian Language

by Milan Mogu. On the basis of this corpus the rst

and in 2002 by Development of the Croatian Language

Croatian frequency dictionary was produced [48]. e

Resources. In 2007 three main research programmes

collection of the Croatian National Corpus [49] started

oriented to the development of LT for Croatian, encom-

in 1998 [50, 20] and it reached 101 million in 2004

passing several research projects were funded from the

[51]. Today, the largest Croatian corpus is the hrWaC

same source:

collected at the same Faculty in 2011 and it reached

1.2 billion tokens crawled from the .hr internet domain

Computational Lingustic Models and Language

[52]. In 2000 at the same Faculty, led by Damir Boras,

Technologies for Croatian [56] where the produc-

a large campain of digitisation of Croatian old mono-

tion and maintaining of a number of resources and

and multilingual dictionaries started [53].

tools has been initiated (e. g., Croatian National

Corpus, Croatian-English Parallel Corpus, Croatian Morphological Lexicon, Croatian Dependency

Did you know that the oldest Croatian printed

dictionary Dictionarium quinque nobilissimarum
Europae linguarum Latinae, Italicae,
Germanicae, Dalmaticae et Ungaricae by Faust
Vrani (1595) is also the oldest Hungarian
printed dictionary?

Treebank [57], Croatian Wordnet [58], hybrid tagger [59] and lemmatiser [15], dependency parser,
NERC system and other information extraction
tools [60] etc.);
Sources for Croatian Heritage and Croatian European Identity [61] with projects dealing with digiti-

At the Institute of Croatian Language and Linguistics

sation of old-Croatian dictionaries and building the

the collection of a comprehensive language corpus e

Croatian valency dictionary [62];

Croatian Language On-line Repository (Riznica) [54,

Croatian Language Repository [54] where a number

55] that includes Croatian written texts from the 11th

of projects deal with dierent linguistic problems

century onward started in 2004. is Repository is or-

starting from Croatian dialects and etymological re-

ganized into three major corpora (Old Croatian, Mid-

search up to the development of semantic networks

dle Croatian, Modern Croatian) where for the rst two

in building lexical resources. ese projects include

a substantial problems characteristic for diachronic cor-

digitisation of collected linguistic data thus enrich-

pora have to be solved, e. g., transliteration of three dif-

ing the pool of available language resources for Croa-

ferent scripts (Glagolitic, Cyrillic and Latin), no stan-

tian.

dardized orthographies, individual variations in the usage of certain characters etc.

Also at the University of Rijeka the project Speech Tech-

Aer the research programmes in 1970s and 1980s, that

nologies [63] made signicant progress in the devel-

were typically oriented to literary and linguistic com-

opment of the basic resources and tools for Croatian

76

speech processing such as Croatian Speech Corpus and

Croatian stands reasonably well with respect to the

prototypes for Croatian ATR and TTS.

most basic language technology tools and resources,

is programmes opened the possibility to catch up

such as reference corpora, smaller parallel corpora,

with the level of LT development in other European

large inectional lexicons, tokenisers, MSD taggers,

languages and enabled the participation of Croatian re-

lemmatisers, NERC system etc.

search teams in current FP7 and ICT PSP projects, since

However, a large syntactically annotated corpus is

the last one that they participated in (TELRI II) n-

missing as well as a large parallel corpus (e. g., Croa-

ished in 2002.

tian translations of Acquis Communautaire). Many

From Croatia the Faculty of Humanites and Social

existing resources lack standardisation so initiatives

Sciences, University of Zagreb was a partner in the

are needed to standardise the data and interchange

CLARIN project a pan-European eort to create a

formats.

language resource infrastructure for researchers in hu-

Experiments have been conducted in some areas,

manities and social sciences and Croatia is to become

such as shallow parsing (chunking), summarization,

one of the member countries of the CLARIN ERIC.

application of ontological resources, but only in an

e same institution takes part in FP7 project AC-

academic research environment. However, the re-

CURAT and ICT-PSP projects LetsMT! and CESAR.

sults obtained are far from the level of development

e University of Zadar was a partner in the ICT-PSP

that other European languages demonstrate. e

project ATLAS.

multimedia and multimodal document processing,

In 2004 the Croatian Language Technologies Society

is gaining attraction, particularly the digitisation in

[69] was founded as a non-governmental organisation

the context of preserving cultural heritage, but lan-

and since then it takes care about the development of

guage technologies for Croatian are not involved in

language technologies for Croatian. e Society has

these processes as needed.

successfuly organised several national as well as interna-

ere exist also individual products with limited

tional conferences, Formal Approaches to South Slavic

functionality in subelds such as speech synthesis,

and Balkan Languages (2008, 2010, 2012) and Slav-

speech recognition and information extraction, and

iCorp (2011), and appeared as a publisher of several

a few others.

books in the eld.

Tools and resources for more advanced language

technology such as deep parsing, machine transla-

4.5 AVAILABILITY OF TOOLS

AND RESOURCES FOR
CROATIAN

Taken the funding of all above mentioned language

Figure 14 provides a rating for language technology sup-

technology programmes and projects from 2007 to

port for Croatian. is rating of existing tools and re-

2012 the amount was only around 1/6 of the estimated

sources was generated by leading experts in the eld who

needed sum. It should therefore come as no surprise that

provided estimates based on a scale from 0 (very low) to

Croatian LT is still in its early stages. 5.5 million speak-

6 (very high) using seven criteria. e key results can be

ers in the Republic of Croatian and neighbouring coun-

summed up as follows:

tries are simply too few to sustain costly development

tion, text semantics, discourse processing, language

generation, dialogue management, etc., simply do
not exist.

77

Coverage

Maturity

Sustainability

Adaptability

Speech synthesis

Grammatical analysis

1.5

3.5

0.3

0.3

0.67

0.3

Text generation Processing

Machine translation

uality

Availability

uantity
Speech recognition

Language Technology: Tools, Technologies, Applications

Semantic analysis

Language Resources: Resources, Data, Knowledge Bases

Text corpora

2.5

Speech corpora

Parallel corpora

Lexical resources

2.5

3.5

3.5

3.5

2.5

2.5

Grammars

14: State of language technology support for Croatian

of new products. At present, almost no companies in

tion areas (machine translation and speech processing)

Croatia are working in the LT area because they do not

and one underlying technology (text analysis), as well

see it as protable. It is thus extremely important to con-

as basic resources needed for building LT applications.

tinue public support for Croatian LT particularly hav-

e languages were categorised using the following ve-

ing in mind the enlargement of digital documents ap-

point scale:

pearing in Croatian since it will become the 24th ocial

language of the European Union by Croatian accession

1. Excellent support

in 2013.

2. Good support
3. Moderate support

4.6 CROSS-LANGUAGE
COMPARISON

4. Fragmentary support

e current state of LT support varies considerably from

LT support was measured according to the following cri-

one language community to another. In order to com-

teria:

pare the situation between languages, this section will

Speech Processing: uality of existing speech recog-

present an evaluation based on two sample applica-

nition technologies, quality of existing speech synthesis

5. Weak or no support

78

technologies, coverage of domains, number and size of

nology community and its related stakeholders are now

existing speech corpora, amount and variety of available

in a position to design a large scale research and develop-

speech-based applications.

ment programme aimed at building a truly multilingual,

Machine Translation: uality of existing MT tech-

technology-enabled communication across Europe.

nologies, number of language pairs covered, coverage of

linguistic phenomena and domains, quality and size of
existing parallel corpora, amount and variety of available
MT applications.
Text Analysis: uality and coverage of existing text
analysis technologies (morphology, syntax, semantics),
coverage of linguistic phenomena and domains, amount
and variety of available applications, quality and size of
existing (annotated) text corpora, quality and coverage
of existing lexical resources (e. g., WordNet) and grammars.
Resources: uality and size of existing text corpora,
speech corpora and parallel corpora, quality and coverage of existing lexical resources and grammars.

e results of this white paper series show that there is a

dramatic dierence in language technology support between the various European languages. While there are
good quality soware and resources available for some
languages and application areas, others, usually smaller
languages, have substantial gaps. Many languages lack
basic technologies for text analysis and the essential resources. Others have basic tools and resources but the
implementation of for example semantic methods is still
far away. erefore a large-scale eort is needed to attain the ambitious goal of providing high-quality language technology support for all European languages,
for example through high quality machine translation.

Figures 15 to 18 show that Croatian is in the bottom

We cannot really be optimistic about technology sup-

cluster for almost all of the tools and resources listed. It

port for the Croatian language. ere is a nascent re-

compares well with other languages with a small num-

search scene in Croatia concerning Croatian language

ber of speakers, such as Estonian, Latvian, Lithuanian,

LT, mostly in universities and scientic institutions, but

Slovak, and to some extent more developed Danish and

the small and medium enterprises are only potential

Finnish. However, all these languages lag far behind

users of solutions of specic LT problems and no de-

large languages like German and French, for instance.

velopment is done there. Various institutions have de-

But even LT resources and tools for those languages

voted their eorts to research and development of the

clearly do not yet reach the quality and coverage of com-

LT products such as production of large Croatian cor-

parable resources and tools for the English language,

pora, the morphology processing, machine translation,

which is in the lead in almost all LT areas. And there are

speech recognition system, etc. But those must be fur-

still plenty of gaps in English language resources with

ther developed and supported. According to the assess-

regard to high quality applications.

ment detailed in this report, immediate action must occur before any breakthroughs for the Croatian language

4.7 CONCLUSIONS

can be achieved. It is clear that there must be a greater

eort to create LT resources for Croatian, and drive re-

In this series of white papers, we have made an impor-

search, innovation and development in general. e

tant eort by assessing the language technology support

need for large amounts of data and the extreme com-

for 30 European languages, and by providing a high-

plexity of language technology systems makes it vital to

leel comparison across these languages. By identifying

develop a new infrastructure to spur greater sharing and

the gaps, needs and decits, the European language tech-

cooperation.

79

Public funding for LT in Europe is relatively low com-

as a partnership between government, academia and in-

pared to the expenditures for language translation and

dustry to develop an expertise cluster in Croatian lan-

multilingual information access by the USA [64]. In

guage technology. We believe that this initiative should

Croatia public funding is even lower than in many other

be institutionally supported by a special-purpose com-

European countries, including neighboring countries

petence centre that could be funded by the EU in order

Slovenia and Hungary. Finally there is a lack of conti-

to stimulate business research and promote sectoral co-

nuity in research and development funding. Short-term

operation between companies and research institutions

coordinated programmes tend to alternate with periods

to develop innovative products and technologies to im-

of sparse or zero funding. In addition, there is an over-

prove the competitiveness of enterprises on the EU mar-

all lack of coordination with programmes in other EU

ket from 2013 on.

countries and at the European Commission level.

e long term goal of META-NET is to enable the cre-

Although there is a pressing need of recognising the

ation of high-quality language technology for all lan-

importance of LT in ensuring sustainable development

guages. is requires all stakeholders in politics, re-

of Croatian in the 21st century and in challenges that

search, business, and society to unite their eorts.

EU membership will bring with the role of Croatian as

e resulting technology will help tear down existing

the 24th EU ocial language, no national initiative has

barriers and build bridges between Europes languages,

been launched, that would foster the creation of large-

paving the way for political and economic unity through

scale resources and tools/services for Croatian, as well

cultural diversity.

80

Excellent
support

Good
support
English

Moderate
support
Czech
Dutch
Finnish
French
German
Italian
Portuguese
Spanish

Fragmentary
support
Basque
Bulgarian
Catalan
Danish
Estonian
Galician
Greek
Hungarian
Irish
Norwegian
Polish
Serbian
Slovak
Slovene
Swedish

Weak/no
support
Croatian
Icelandic
Latvian
Lithuanian
Maltese
Romanian

15: Speech processing: state of language technology support for 30 European languages

Excellent
support

Good
support
English

Moderate
support
French
Spanish

Fragmentary
support
Catalan
Dutch
German
Hungarian
Italian
Polish
Romanian

Weak/no
support
Basque
Bulgarian
Croatian
Czech
Danish
Estonian
Finnish
Galician
Greek
Icelandic
Irish
Latvian
Lithuanian
Maltese
Norwegian
Portuguese
Serbian
Slovak
Slovene
Swedish

16: Machine translation: state of language technology support for 30 European languages

81

Excellent
support

Good
support
English

Moderate
support
Dutch
French
German
Italian
Spanish

Fragmentary
support
Basque
Bulgarian
Catalan
Czech
Danish
Finnish
Galician
Greek
Hungarian
Norwegian
Polish
Portuguese
Romanian
Slovak
Slovene
Swedish

Weak/no
support
Croatian
Estonian
Icelandic
Irish
Latvian
Lithuanian
Maltese
Serbian

17: Text analysis: state of language technology support for 30 European languages

Excellent
support

Good
support
English

Moderate
support
Czech
Dutch
French
German
Hungarian
Italian
Polish
Spanish
Swedish

Fragmentary
support
Basque
Bulgarian
Catalan
Croatian
Danish
Estonian
Finnish
Galician
Greek
Norwegian
Portuguese
Romanian
Serbian
Slovak
Slovene

Weak/no
support
Icelandic
Irish
Latvian
Lithuanian
Maltese

18: Speech and text resources: State of support for 30 European languages

82

5
ABOUT META-NET
META-NET is a Network of Excellence partially

e main focus of this activity is to build a coherent

funded by the European Commission. e network

and cohesive LT community in Europe by bringing to-

currently consists of 54 research centres members from

gether representatives from highly fragmented and di-

33 European countries. META-NET forges META, the

verse groups of stakeholders. e present White Paper

Multilingual Europe Technology Alliance, a growing

was prepared together with volumes for 29 other lan-

community of language technology professionals and

guages. e shared technology vision was developed in

organisations in Europe [65].

three sectorial Vision Groups. e META Technology

META-NET fosters the technological foundations for a

Council was established in order to discuss and to pre-

truly multilingual European information society that:

pare the SRA based on the vision in close interaction

makes communication and cooperation possible

with the entire LT community.

across languages;
grants all Europeans equal access to information and
knowledge in any language;
builds upon and advances functionalities of networked information technology.
e network supports a Europe that unites as a single digital market and information space. It stimulates
and promotes multilingual technologies for all European languages. ese technologies support automatic

META-SHARE creates an open, distributed facility

for exchanging and sharing resources.

e peerto-

peer network of repositories will contain language data,

tools and web services that are documented with highquality metadata and organised in standardised categories. e resources can be readily accessed and uniformly searched. e available resources include free,
open source materials as well as restricted, commercially
available, fee-based items.

translation, content production, information process-

META-RESEARCH builds bridges to related technol-

ing and knowledge management for a wide variety of

ogy elds. is activity seeks to leverage advances in

subject domains and applications. ey also enable in-

other elds and to capitalise on innovative research that

tuitive language-based interfaces to technology rang-

can benet language technology. In particular, the ac-

ing from household electronics, machinery and vehi-

tion line focuses on conducting leading-edge research in

cles to computers and robots. Launched on 1 February

machine translation, collecting data, preparing data sets

2010, META-NET has already conducted various activ-

and organising language resources for evaluation pur-

ities in its three lines of action META-VISION, META-

poses; compiling inventories of tools and methods; and

SHARE and METARESEARCH.

organising workshops and training events for members

META-VISION fosters a dynamic and inuential

of the community.

stakeholder community that unites around a shared vision and a common strategic research agenda (SRA).

oce@meta-net.eu http://www.meta-net.eu

83

A
BIBLIOGRAFIJA REFERENCES
[1] Aljoscha Burchardt, Markus Egg, Kathrin Eichler, Brigitte Krenn, Jrn Kreutel, Annette Lemllmann, Georg Rehm,
Manfred Stede, Hans Uszkoreit, and Martin Volk. Die Deutsche Sprache im Digitalen Zeitalter e German Language
in the Digital Age. META-NET White Paper Series. Georg Rehm and Hans Uszkoreit (Series Editors). Springer, 2012.
[2] Aljoscha Burchardt, Georg Rehm, and Felix Sasaki. e Future European Multilingual Information Society Vision
Paper for a Strategic Research Agenda, 2011. http://www.meta-net.eu/vision/reports/meta-net-vision-paper.pdf.
[3] Directorate-General Information Society & Media of the European Commission. User Language Preferences Online,
2011. http://ec.europa.eu/public_opinion/flash/fl_313_en.pdf.
[4] European Commission. Multilingualism: an Asset for Europe and a Shared Commitment, 2008. http://ec.europa.eu/
languages/pdf/comm2008_en.pdf.
[5] Directorate-General of the UNESCO. Intersectoral Mid-term Strategy on Languages and Multilingualism, 2007. http:
//unesdoc.unesco.org/images/0015/001503/150335e.pdf.
[6] Directorate-General for Translation of the European Commission. Size of the Language Industry in the EU, 2009.
http://ec.europa.eu/dgs/translation/publications/studies.
[7] Narodne novine. Ustav Republike Hrvatske, 2001. http://narodne-novine.nn.hr/clanci/sluzbeni/232289.html.
[8] Mladen Klemeni. A Concise atlas of the Republic of Croatia & of the Republic of Bosnia and Hercegovina. Miroslav
Krlea Lexicographical Institute, 1993.
[9] Institut za hrvatski jezik i jezikoslovlje (Institute of Croatian Language and Linguistics). http://www.ihjj.hr.
[10] Institut za hrvatski jezik i jezikoslovlje. Jezini Savjeti (Language Advice Portal). http://savjetnik.ihjj.hr.
[11] Institut za hrvatski jezik i jezikoslovlje. Struna: Hrvatsko strukovno nazivlje (Struna: Croatian Professional Terminology). http://struna.ihjj.hr/o-programu/.
[12] Croaticum centar za hrvatski kao drugi i strani jezik (Croaticum e Center for Croatian as a Second and Foreign
Language). http://croaticum.ffzg.hr.
[13] Hrvatski jezik (Croatian Language). http://www.hrvatskijezik.eu.
[14] Jezine tehnologije za hrvatski jezik (HLT). http://jthj.ffzg.hr.
[15] eljko Agi, Marko Tadi, and Zdravko Dovedan. Evaluating Full Lemmatization of Croatian Texts. In M. Klopotek,
A. Przepiorkowski, S. Wierzchon, and K. Trojanowski, editors, Recent Advances in Intelligent Information Systems. Academic Publishing House EXIT, 2009.
[16] Daniel Jurafsky and James H. Martin. Speech and Language Processing (2nd Edition). Prentice Hall, 2009.

85

[17] Christopher D. Manning and Hinrich Schtze. Foundations of Statistical Natural Language Processing. MIT Press,
1999.
[18] Language Technology World (LT World). http://www.lt-world.org.
[19] Ronald Cole, Joseph Mariani, Hans Uszkoreit, Giovanni Battista Varile, Annie Zaenen, and Antonio Zampolli, editors.
Survey of the State of the Art in Human Language Technology (Studies in Natural Language Processing). Cambridge
University Press, 1998.
[20] Marko Tadi. Jezine tehnologije i hrvatski jezik (HLT and Croatian). Exlibris, 2003.
[21] Hrvatski akademski spelling checker (Hascheck). http://hacheck.tel.fer.hr.
[22] Spiegel Online. Google zieht weiter davon (Google is still leaving everybody behind), 2009. http://www.spiegel.de/
netzwelt/web/0,1518,619398,00.html.
[23] Juan Carlos Perez. Google Rolls out Semantic Search Capabilities, 2009. http://www.pcworld.com/businesscenter/
article/161869/google_rolls_out_semantic_search_capabilities.html.
[24] Hrvatski Morfoloki Leksikon (Croatian Morphological Lexicon). http://hml.ffzg.hr.
[25] MULTEXT-East: Multilingual Text Tools and Corpora for Central and Eastern European Languages. http://nl.ijs.si/
ME/.
[26] Cadial: Computer aided document indexing for accessing legislation. http://www.cadial.org.
[27] Cadial: Computer aided document indexing for accessing legislation. http://cadial.hidra.hr/search.php.
[28] e MBROLA Project. http://tcts.fpms.ac.be/synthesis/mbrola.html.
[29] Branimir Dropulji and Davor Petrinovi. Development of Acoustic Model for Croatian Language Using HTK. Automatika, 51(1):7988, 2010.
[30] Sanda Martini-Ipi, Miran Pobar, and Ivo Ipi. Croatian Large Vocabulary Automatic Speech Recognition. Automatika, 52(2):147157, 2011.
[31] CRO-SPEECHDAT (Baza govornih uzoraka i tekstova dostupna putem Interneta). http://www.inf.uniri.hr/~ivoi/
CROSPEECH/index.htm.
[32] Sanda Martini-Ipi and Ivo Ipi. Croatian HMM Based Speech Synthesis. Journal of Computing and Information
Technology, 14(4):299305, 2006.
[33] Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. BLEU: a Method for Automatic Evaluation of Machine
Translation. In Proceedings of the 40th Annual Meeting of ACL, 2002.
[34] Philipp Koehn, Alexandra Birch, and Ralf Steinberger. 462 Machine Translation Systems for Europe. In MT Summit
XII, 2009.
[35] Svetozar Petrovi and Bulcs Lszl. Strojno prevoenje i statistika u jeziku. Nae teme, (6):105298, 1959.
[36] Information Technology in Translation and e-Learning of Croatian. http://rmjt.ffzg.hr/p4_en.html.
[37] Lets MT. https://www.letsmt.eu/Start.aspx.
[38] Accurat. http://www.accurat-project.eu.

86

[39] Inguna Skadia, Andrejs, Vasijevs Raivis Skadi, Robert Gaizauskas, Dan Tu, and Tatiana Gornostay. Analysis and
Evaluation of Comparable Corpora for Under Resourced Areas of Machine Translation. In Proceedings of the 3rd Workshop on Building and Using Comparable Corpora. European Language Resources Association (ELRA), 2010.
[40] Andreas Eisele and Jia Xu. Improving Machine Translation Performance Using Comparable Corpora. In Proceedings of
the 3rd Workshop on Building and Using Comparable Corpora, 2010.
[41] Andrejs Vasiljevs, Tatiana Gornostay, and Raivis Skadins. LetsMT! Online Platform for Sharing Training Data and
Building User Tailored Machine Translation. In Proceedings of the Fourth Baltic conference Human Language Technologies the Baltic Perspective, 2010.
[42] HINA: Hrvatska izvjetajna novinska agencija (HINA: Croatian News Agency). http://websrv2.hina.hr/hina/web/
index.action.
[43] Knowledge Technologies Lab. http://ktlab.fer.hr.
[44] Nives Mikeli Preradovi, Tomislava Lauc, and Damir Boras. CROXMLSUM the System for XML. Document Summarization in Croatian. International Journal of Mathematics and Computers in Simulation, 1(1):8189, 2007.
[45] Branko itko, Slavomir Stankov, Marko Rosi, and Ani Grubii. Dynamic test generation over ontology-based knowledge representation in authoring shell. Expert Systems with Applications, 36(4):81858196, 2009.
[46] eljko Bujas. Osman, kompjutorska konkordancija (Osman, Computer Concordance). Sveuilina naklada Liber, 1974.
[47] Marko Tadi. Raunalna obradba hrvatskih korpusa: povijest, stanje i perspektive (Computer processing of Croatian
corpora: history, status and perspectives). Suvremena lingistika, 43-44(1-3):387394, 1997.
[48] Milan Mogu, Maja Bratani, and Marko Tadi. Hrvatski estotni rjenik (Croatian Frequency Dictionary). kolska
knjiga, 1999.
[49] Hrvatski nacionalni korpus (Croatian National Corpus). http://hnk.ffzg.hr.
[50] Marko Tadi. Building the Croatian National Corpus. In Proceedings of the 3rd International Conference on Language
Resources and Evaluation (LREC2002), 2002.
[51] Marko Tadi. New version of the Croatian National Corpus. In Dana Hlavkov, Ale Hork, Klara Osolsob, and Pavel
Rychl, editors, Aer Half a Century of Slaonic Natural Language Processing, Masaryk Uniersity. Masaryk University,
2009.
[52] Nikola Ljubei and Toma Erjavec. hrWaC and slWaC: Compiling web corpora for Croatian and Slovene. In Proceedings of the 14th International Conference Text, Speech and Dialogue (TSD2011). Springer, 2011.
[53] Portal hrvatske rjenike batine (Croatian Old Dictionary Portal). http://crodip.ffzg.hr.
[54] Hrvatska jezina riznica (Croatian Language Repository). http://riznica.ihjj.hr.
[55] Dunja Brozovi Ronevi and Damir avar. Hrvatska jezina riznica kao podloga jezinim i jezinopovijesnim istraivanjima hrvatskoga jezika ( Croatian Language treasury as a base language and ..... Croatian language studies). In Vidjeti
Ohrid: Proceedings of the 14th international Slaistic Congress in Ohrid, 2008.
[56] Raunalnolingvistiki modeli i jezine tehnologije za hrvatski jezik (Computational Linguistic Models and Language
Technologies for Croatian). http://rmjt.ffzg.hr.
[57] Hrvatska ovisnosna banka stabala (Croatian Dependency Treebank). http://hobs.ffzg.hr.

87

[58] Lexical Semantics in Building the Croatian WordNet. http://rmjt.ffzg.hr/p3_en.html.

[59] eljko Agi, Marko Tadi, and Zdravko Dovedan. Improving Part-of-Speech Tagging Accuracy for Croatian by Morphological Analysis. Informatica, 32(4):445451, 2008.
[60] Knowledge discovery in textual data. http://rmjt.ffzg.hr/p5_en.html.
[61] Ministarstvo znanosti, obrazovanja i sporta. Z projekti. http://zprojekti.mzos.hr/page.aspx?pid=97&lid=1.
[62] Nives Mikeli Preradovi. CROVALLEX lexicon improvements: Subcategorization and semantic constraints. WSEAS
Transactions on Computers, 9(3), 2010.
[63] obrazovanja i sporta Ministarstvo znanosti. Z projekti. http://zprojekti.mzos.hr/page.aspx?pid=96.
[64] Gianni Lazzari. Sprachtechnologien fr Europa, 2006. http://tcstar.org/pubblicazioni/D17_HLT_DE.pdf.
[65] Georg Rehm and Hans Uszkoreit. Multilingual Europe: A challenge for language tech. MultiLingual, 22(3):5152,
April/May 2011.
[66] Jerrold H. Zar. Candidate for a Pullet Surprise. Journal of Irreproducible Results, page 13, 1994.
[67] Ralf Steinberger, Bruno, Pouliquen, Anna Widiger, Camelia Ignat, Toma Erjavec, Dan Tu, and Dniel Varga. e
JRC-Acquis: A multilingual aligned parallel corpus with 20+ languages. In Proceedings of the 5th International Conference on Language Resources and Evaluation (LREC2006), 2006.
[68] Boo Bekavac and Marko Tadi. Implementation of Croatian NERC system. In Proceedings of the Workshop on BaltoSlaonic Natural Language Processing 2007, 2007.
[69] Hrvatsko drutvo za jezine tehnologije (Croatian LT Society). http://www.hdjt.hr/index_en.html.

88

B
META-NET LANICE META-NET MEMBERS
Austrija

Austria

Zentrum fr Translationswissenscha, Universitt Wien: Gerhard Budin

Belgija

Belgium

Computational Linguistics and Psycholinguistics Research Centre, University of

Antwerp: Walter Daelemans
Centre for Processing Speech and Images, University of Leuven: Dirk van Compernolle

Bugarska

Bulgaria

Institute for Bulgarian Language, Bulgarian Academy of Sciences: Svetla Koeva

Cipar

Cyprus

Language Centre, School of Humanities: Jack Burston

eka

Czech Republic

Institute of Formal and Applied Linguistics, Charles University in Prague: Jan Haji

Danska

Denmark

Centre for Language Technology, University of Copenhagen:

Bolette Sandford Pedersen, Bente Maegaard

Estonija

Estonia

Institute of Computer Science, University of Tartu: Tiit Roosmaa, Kadri Vider

Finska

Finland

Computational Cognitive Systems Research Group, Aalto University: Timo Honkela

Department of Modern Languages, University of Helsinki:
Kimmo Koskenniemi, Krister Lindn

Francuska

France

Centre National de la Recherche Scientique, Laboratoire dInformatique pour la Mcanique et les Sciences de lIngnieur and Institute for Multilingual and Multimedia Information: Joseph Mariani
Evaluations and Language Resources Distribution Agency: Khalid Choukri

Grka

Greece

R.C. Athena, Institute for Language and Speech Processing: Stelios Piperidis

Hrvatska

Croatia

Institute of Linguistics, Faculty of Humanities and Social Science, University of Zagreb:

Marko Tadi

Irska

Ireland

School of Computing, Dublin City University: Josef van Genabith

Island

Iceland

School of Humanities, University of Iceland: Eirkur Rgnvaldsson

Italija

Italy

Consiglio Nazionale delle Ricerche, Istituto di Linguistica Computazionale

Antonio Zampolli: Nicoletta Calzolari
Human Lang. Technology, Fondazione Bruno Kessler: Bernardo Magnini

Latvija

Latvia

Tilde: Andrejs Vasijevs

Institute of Mathematics and Computer Science, University of Latvia: Inguna Skadia

Litva

Lithuania

Institute of the Lithuanian Language: Jolanta Zabarskait

Luksemburg

Luxembourg

Arax Ltd.: Vartkes Goetcherian

89

Maarska

Hungary

Research Institute for Linguistics, Hungarian Academy of Sciences: Tams Vradi

Department of Telecommunications and Media Informatics, Budapest University of
Technology and Economics: Gza Nmeth, Gbor Olaszy

Malta

Malta

Department Intelligent Computer Systems, University of Malta: Mike Rosner

Nizozemska

Netherlands

Utrecht Institute of Linguistics, Utrecht University: Jan Odijk

Computational Linguistics, University of Groningen: Gertjan van Noord

Norveka

Norway

Department of Linguistic, Literary and Aesthetic Studies, University of Bergen:

Koenraad De Smedt
Department of Informatics, Language Technology Group, University of Oslo:
Stephan Oepen

Njemaka

Germany

Language Technology Lab, DFKI: Hans Uszkoreit, Georg Rehm

Human Language Technology and Pattern Recognition, RWTH Aachen University:
Hermann Ney
Department of Computational Linguistics, Saarland University: Manfred Pinkal

Poljska

Poland

Institute of Computer Science, Polish Academy of Sciences:

Adam Przepirkowski, Maciej Ogrodniczuk
University of d: Barbara Lewandowska-Tomaszczyk, Piotr Pzik
Department of Computer Linguistics and Articial Intelligence, Adam Mickiewicz University: Zygmunt Vetulani

Portugal

Portugal

University of Lisbon: Antnio Branco, Amlia Mendes

Spoken Language Systems Laboratory, Institute for Systems Engineering and Computers: Isabel Trancoso

Rumunjska

Romania

Research Institute for Articial Intelligence, Romanian Academy of Sciences:

Dan Tu
Faculty of Computer Science, University Alexandru Ioan Cuza of Iai: Dan Cristea

Slovaka

Slovakia

udovt tr Institute of Linguistics, Slovak Academy of Sciences: Radovan Garabk

Slovenija

Slovenia

Joef Stefan Institute: Marko Grobelnik

Srbija

Serbia

University of Belgrade, Faculty of Mathematics: Duko Vitas, Cvetana Krstev,

Ivan Obradovi
Pupin Institute: Sanja Vranes

panjolska

Spain

Barcelona Media: Toni Badia, Maite Melero

Institut Universitari de Lingstica Aplicada, Universitat Pompeu Fabra: Nria Bel
Aholab Signal Processing Laboratory, University of the Basque Country:
Inma Hernaez Rioja

90

Center for Language and Speech Technologies and Applications, Universitat Politcnica
de Catalunya: Asuncin Moreno
Department of Signal Processing and Communications, University of Vigo:
Carmen Garca Mateo
vedska

Sweden

Department of Swedish, University of Gothenburg: Lars Borin

vicarska

Switzerland

Idiap Research Institute: Herv Bourlard

UK

UK

School of Computer Science, University of Manchester: Sophia Ananiadou

Institute for Language, Cognition and Computation, Center for Speech Technology Research, University of Edinburgh: Steve Renals
Research Institute of Informatics and Language Processing, University of Wolverhampton: Ruslan Mitkov

Oko 100 jezine tehnologije strunjaci Predstavnici zemalja i jezika zastupljenih u META-NET raspravlja
i nalizirani kljune rezultate i poruke Bijele knjige serije na sastanku u Berlinu, Njemaka, listopada 21/22,
2011. About 100 language technology experts representatives of the countries and languages represented
in META-NET discussed and nalised the key results and messages of the White Paper Series at a meeting in
Berlin, Germany, on October 21/22, 2011.

91

C
NIZ BIJELE THE META-NET
KNJIGE META-NET WHITE PAPER SERIES
baskijski
bugarski
eki
danski
engleski
estonski
nski
francuski
galicijski
grki
hrvatski
irski
islandski
katalonski
latvijski
litavski
maarski
malteki
nizozemski
norveki bokml
norveki nnorsk
njemaki
poljski
portugalski
rumunjski
slovaki
slovenski
srpski
panjolski
vedski
talijanski

Basque
Bulgarian
Czech
Danish
English
Estonian
Finnish
French
Galician
Greek
Croatian
Irish
Icelandic
Catalan
Latvian
Lithuanian
Hungarian
Maltese
Dutch
Norwegian Bokml
Norwegian Nynorsk
German
Polish
Portuguese
Romanian
Slovak
Slovene
Serbian
Spanish
Swedish
Italian

euskara

etina
dansk
English
eesti
suomi
franais
galego

hrvatski
Gaeilge
slenska
catal
latvieu valoda
lietuvi kalba
magyar
Malti
Nederlands
bokml
nynorsk
Deutsch
polski
portugus
romn
slovenina
slovenina

espaol
svenska
italiano

93

Research
Co

ies
unit
mm

Lan
gu
a

es
stri
u
d

Soc
iet

rs
Use
e
g

In

In everyday communication, Europes citizens, business

U svakodnevnoj komunikaciji graani Europe, pos-

partners and politicians are inevitably confronted with

lovni partneri i politiari neizbjeno su suoeni s je-

language barriers. Language technology has the po-

zinim barijerama. Potencijal koji imaju jezine teh-

tential to overcome these barriers and to provide inno-

nologije mogao bi savladati te prepreke i osigurati

vative interfaces to technologies and knowledge. This

inovativna suelja za tehnologije i znanja. Ovaj do-

white paper presents the state of language technology

kument prikazuje stanje jezinih tehnologija za hr-

support for the Icelandic language. It is part of a se-

vatski jezik. Jedan je od dokumenata u nizu bijele

ries that analyses the available language resources and

knjige koji analizira dostupne jezine resurse i tehno-

technologies for 30 European languages. The analysis

logije za 30 europski jezik. Analizu je proveo META-

was carried out by META-NET, a Network of Excellence

NET mrea izvrsnosti koju nancira Europska komi-

funded by the European Commission. META-NET con-

sija. META-NET se sastoji od 54 istraivaka centra u

sists of 54 research centres in 33 countries, who cooper-

33 zemalje, koji surauju s partnerima iz gospodar-

ate with stakeholders from economy, government agen-

stva, dravnih agencija, istraivakih organizacija i

cies, research organisations and others. META-NETs

drugih nevladinih organizacija, jezinih zajednica i

vision is high-quality language technology for all Euro-

europskih sveuilita. Vizija je META-NET-a povea-

pean languages.

nje kvalitete jezinih tehnologija za sve europske jezike.

Niz Jezinih bijelih knjiga otvara nove uvide u europsku jezinu raznolikost dok istodobno relativizira pojam
tzv. malih jezika, poput hrvatskoga. Stoga jezine tehnologije imaju ne samo kljunu ulogu u iskazivanju jezinoga bogatstva u dananjoj Europi, ve predstavljaju metodoloko ishodite za daljnji razvitak digitalnih humanistikih znanosti, osobito ako ih se promatra kao temelj za daljnja istraivanja u raznim humanistikih disciplinama.
Prof. dr. Milena ic Fuchs (redoviti lan Hrvatske akademije znanosti i umjetnosti, predsjednica Stalnoga odbora
za humanistike znanosti Europske znanstvene zaklade)

www.meta-net.eu
www.meta-net.eu