 S aipsniai / A icles 223
NATALIIADARCHUK
Ta asShe chenkoNa ionalUni e si yo Kyi
ORCIDid:o cid.o g/0000-0001-8932-9301
Fieldso  esea ch:compu e linguis ics,co puslinguis ics,
quan i a i elinguis ics,g amma andseman icso  he
Uk ainianlanguage.
OKSANAZUBAN
Ta asShe chenkoNa ionalUni e si yo Kyi
ORCIDid:o cid.o g/0000-0002-2644-3892
Fieldso  esea ch:compu e linguis ics,linguis icexpe ise,
quan i a i elinguis ics,g amma andseman icso  he
Uk ainianlanguage.
VALENTYNAROBEIKO
Ta asShe chenkoNa ionalUni e si yo Kyi
ORCIDid:o cid.o g/0000-0003-2266-7650
Fieldso  esea ch:speechanalysis, ecogni ionand
syn hesis,phone ics,na u allanguagep ocessing.
YULIIATSYHVINTSEVA
Ta asShe chenkoNa ionalUni e si yo Kyi ,
Ins i u eo  heUk ainianLanguageo  heNa ional
Academyo Scienceso Uk aine
ORCIDid:o cid.o g/0000-0002-9684-3840
Fieldso  esea ch:Uk ainianneology,lexicologyand
lexicog aphy.
VICTORSOROKIN
Ta asShe chenkoNa ionalUni e si yo Kyi
ORCIDid:o cid.o g/0000-0002-3637-0535
Fieldso  esea ch:au oma icsyn axanalysis,au oma ic
seman icanalysis,na u allanguagep ocessing.
NATALIIA DARCHUK, OKSANA ZUBAN, VALENTYNA ROBEIKO,
YULIIA TSYHVINTSEVA, VICTOR SOROKIN, MYKOLA SAZHOK
224 Ac aLinguis icaLi huanicaXCI
MYKOLASAZHOK
Ins i u e o in o ma ion echnologiesandsys ems
o  heNa ionalAcademyo Scienceso Uk aine
ORCIDid:o cid.o g/0000-0003-1169-6851
Fieldso  esea ch:speechanalysis, ecogni ionand
syn hesis,na u allanguagep ocessing.
DOI:doi.o g/10.35321/all91-09
THESYSTEMFORAUTOMATIC
STYLOMETRICANALYSIS
OFUKRAINIANMEDIA
TEXTSTEXTATTRIBUTOR1.0
(TECHNIQUES,MEANS,
FUNCTIONALITY)
Uk ainosžiniasklaidos eks ųau oma inės
s ilome inėsanalizėssis ema„Tex A ibu o 1.0“
(me odai,p iemonės, unkcionalumas)
ANNOTATION
Thispape p esen s hes uc u e,algo i hms,implemen a ion,andexpe imen al esul s
o  heau oma icTex A ibu o sys emde elopedby heau ho so  hepape  o s a is ical
Uk ainian-language ex pa ame e isa ionusingamul ipa ame icse o s a is icalindices,
cha ac e ising heau ho ’s ex s yleandapplicable oau ho shipa ibu ion asks.Based
on hec ea edlinguis ic esou cesandso wa e, hesys emgene a esalinguis icanalysis
basedon hecalcula eds a is icalindicesandpe o msacompa a i es udyo  wo ex s.An
addi ionalc i e ion o s a is icalindexingis he ex  oxici yindex,calcula ed h ough he
me hodo  e baliden i ica iono  oxicsen imen .Au ho shipand oxici yde ec ion asks
a eadd essedusing wome hods:dic iona y-and ule-baseds a is icalcalcula ionsand
machinelea ning.Thecu en  indingsimplemen edin hebe a e siono Tex A ibu o 
a e ho oughlyexamined.
 KEYWORDS: Compu a ionallinguis ics,Uk ainianlanguage,sen imen analysis,
au ho shipa ibu ion,s ylome y, ex classi ica ion.
 S aipsniai / A icles 225
The Sys em o Au oma ic S ylome ic Analysis
o Uk ainian Media Tex s Tex A ibu o 1.0
(Techniques, Means, Func ionali y)
ANOTACIJA
Šiame s aipsnyje p is a oma s aipsnio au o ių suku os au oma inės sis emos
„Tex A ibu o “,ski oss a is iniamuk ainiečiųkalbos eks ųpa ame iza imuinaudojan 
daugiapa ame inįs a is inių odiklių inkinį,apibūdinan įau o iaus eks os iliųi  aikomą
au o ys ės a ibucijos užda iniams, s uk ū a, algo i mai, įdiegimas i  ekspe imen iniai
ezul a ai.Suku ųling is iniųiš ekliųi p og aminėsį angospag indusis emagene uoja
ling is inęanalizępagalapskaičiuo uss a is iniusindeksusi a liekad iejų eks ųlyginamąją
analizę.Papildomass a is inioindeksa imok i e ijusy aneigiamasnuo aikassukeliančio
eks o indeksas, apskaičiuojamas aikan  žodinio neigiamų nuo aikų iden i ika imo
me odą.Au o ys ėsi pagiežosnus a ymoužduo yssp endžiamosd iemme odais:žodynu
i  aisyklėmis pag įs ais s a is iniais skaičia imais i  mašininiu mokymusi. Daba iniai
ezul a ai,gau inaudojan is„Tex A ibu o “be a e sija,išsamiaiišnag inė i.
 ESMINIAIŽODŽIAI:kompiu e inėling is ika,uk ainiečiųkalba,jausmingumoanalizė,
au o ys ėsp isky imas,s ilome ija, eks oklasi ika imas.
1. INTRODUCTION
Tex ological esea chgainednewscien i icimpo ancewi h head ancemen 
o compu a ionallinguis icsme hods,whichcon ibu ed o he o ma iono a
newdi ec ion ha canbeconside eddigi al ex ology.Wi hin hiseme ging
discipline,weconcep ualizedigi al ex ologyasha nessingau oma edco pus
linguis ics echniques,coupledwi hma hema icalmodels o quan i a i e ex 
analysis.Guidedby hecu en  askso mode nquan i a i elinguis icsand
na u allanguagep ocessing,ou  eamhasde elopedasys em o au oma ic
linguis ic-s a is ical analysis o  media ex s, which has been implemen ed
as a web applica ion named Tex A ibu o  (Tex A ibu o  1.0 2024). The
sys emope a esin ou  asks:1)s a is ical ex pa ame e iza ion;2)s ylome y:
de e mining he linguis ic-s a is ical ea u es o  idiolec ; 3) a ibu ion:
de e mining hedeg eeo simila i ybe ween ex s;and4)sen imen analysis:
iden i ying nega i e sen imen  lexicon in ex s. Addi ionally, wi hin he
sys em, wo linguis ic expe  conclusions a e au oma ically gene a ed based
on he esul so  hesecondand ou h asks.Themul i unc ionali yo  he
Tex A ibu o sys em equi esuse s o amilia ize hemsel eswi hi sope a ing
p inciples.Thepu poseo  hisa icleis oacquain  heacademicandeduca ional
philologicalcommuni ywi h hesys em’sc ea ionme hodology,a chi ec u e,
and ope a ional esul s o p o ide a clea  unde s anding o  i s unc ions
and analy ical capabili ies, which a e pa icula ly ele an . This app oach is
especially ele an  o analyzingla ge olumeso in e ne communica ionsand
NATALIIA DARCHUK, OKSANA ZUBAN, VALENTYNA ROBEIKO,
YULIIA TSYHVINTSEVA, VICTOR SOROKIN, MYKOLA SAZHOK
226 Ac aLinguis icaLi huanicaXCI
enhancingin o ma ionde ences a egiesdu ing heongoingRussian-Uk ainian
wa .Theconcep o de eloping anau oma icsys emo Uk ainian-language
ex a ibu iona ose om heanalysiso  esea chin he ieldo Uk ainian
co puslinguis ics(Da čuk 2013)ands a is icals udieso  heau ho ’s s yle
(Da čuke al.2021;Da chuk,So okin2022;Zuban’2019)conduc edusing he
oolso  heUk ainianLanguageCo pus(KUM)by hep ojec au ho s.
Du ing he de elopmen  o  he Tex A ibu o  sys em, he ollowing
me hodswe eemployed:componen analysis,dis ibu i eanalysis,andcon en 
analysis.Con en analysis,aquan i a i eandquali a i eme hod o ex ac ing
in o ma ion om ex , u ilized na u al language p ocessing and s a is ical
echniques.Quan iza ionandsen imen analysiswe ealsoapplied.Sen imen 
analysis au oma ically iden i ies ex  onali y based on bo h he emo ional
colou ingand heau ho ’sassessmen o e en so objec s,whichwasachie ed
u ilizingdic iona y-and ule-baseds a is icalcalcula ions.
Addi ionally, machine lea ning me hods, including deep lea ning, we e
inco po a ed.Thes a is icals uc u eo a ex unde s oodasi squan i a i e
model,enables heiden i ica iono i s unc ionals yle,au ho ship,andpe iod
o c ea ion.In hiswo k, hes a is icals uc u ecomponen swe eex ac edby
analyzinglexicalandg amma icals a is ical ea u esusingindexing echniques
andEuclideandis anceme ics.TheEuclideandis ance,whichmeasu es he
dis ancebe ween wopoin sinann-dimensionalEuclideanspace,isoneo  he
mos widelyusedme icsinlinguos a is ics o clus e analysisins ylome y
andau ho shipa ibu ion.
The no el y o  he esul s is he i s  implemen a ion in compu a ional
linguis ics o  an au oma ed mo pho-syn ac ic-seman ic s ylome ic model
o  ex  analysis, based on 15 s a is ical indices ha  p ima ily pa ame e ize
hemo phological,aswellassyn ac icandseman ics uc u eo  he ex .A
s ylome icmodelisase o s a is icalpa ame e so a ex (lexical,mo phological,
syn ac ic,e c.),basedonwhichacompa a i eanalysiso  his ex isca ied
ou wi h hes ylome icmodelo  he unc ionals yle.Inou opinion,i is
possible o ob ain a s ic , de e minis ic, scien i ically g ounded sys em o 
commonanddis inc i e ea u esins yles,gen es,e c.byusings a icme hods
in hecons uc iono as ylome icmodel.Fo  he i s  ime, hes ylome ic
modelin oduces he oxici yindexpa ame e ,whichcha ac e is icallyde ines
media ex sdu ing heRussian-Uk ainianwa .Thismodelisimplemen edno 
onlyin he unc iono s a is icalpa ame e iza iono  ex s,whichis ypical
o sys emso  his ype,bu alsoin he unc iono compa ing heanalyzed
ex wi h hemedias yleo  heUk ainianlanguage,and o s ylome icand
a ibu ion asks in ol ing wo o  mo e ex s. The heo e ical signi icance
ex ends o he alida ion o  wo me hods o  de e mining ex  oxici y and
 S aipsniai / A icles 227
The Sys em o Au oma ic S ylome ic Analysis
o Uk ainian Media Tex s Tex A ibu o 1.0
(Techniques, Means, Func ionali y)
au ho shipinUk ainian:1) h oughdic iona yand ule-basedapp oaches,and
2) iamachinelea ning,includingdeeplea ning.
2. RELATEDWORKS
Inmode nUk ainianlinguis ics,subs an iallinguis ic-s a is icals udiesha e
beenconduc edon heidiolec so Uk ainianw i e sandpoe s,includingTa as
She chenko, Lesya Uk ainka, Vasyl S us (Zuban’ 2019), I an F anko (Buk
2021),RomanI anychuk(Lo o ska2022),Vale iiShe chuk(Volos,Le chenko
2023), Ma iia Ma ios, Yu ii And ukho ych, Oksana Zabuzhko (Ka aso ,
Le chenko 2024), I an D ach, Mykola Ving ano sky (Da chuk e al. 2024),
LinaKos enko(Zuban’2019;Da chuke al.2024),amongo he s.Theses udies
we econduc edon ep esen a i e ex samples(long ex s)usingp edominan ly
oneo a ew(nomo e han i e)s a is icalpa ame e s,o enwi hou applying
s ylome iccompa ison.Fu he mo e,due o helabo iousna u eo conduc ing
linguis ic-s a is icalexpe imen s,Uk ainianlinguis icshasla gelyo e looked
comp ehensi es a is icalpa ame e iza ion,compa isono analyzed ex swi h
s anda ds a is icalpa ame e so  unc ionals yles,de e mina iono  hedeg ee
o  ex simila i y,aswellass ylome icanda ibu ionanalysiso sho  ex s.
Theuseo  heTex A ibu o sys eminlinguis ic-s a is ical esea ch,aimed
a  pe o ming hese asks au oma ically, will no  only p o ide equency
cha ac e is ics o  linguis ic phenomena bu  also enable e icien  s ylome ic
analysis esul inginanexpe opinion.
The Tex A ibu o  sys em has ad an ages compa ed o i s coun e pa s
inglobalscience.Mos exis ingsys emsandmodelsalso ocuson heuseo 
indi iduallinguis ic-s a is icalmodules oiden i yoneo a ew,mos ly o mal
(n-g ams,mos  equen wo ds,le e s), ex pa ame e s(Ede 2015;Canhasie al.
2022;LIWC-22;Khomy skae al.2023;ALIAS).Incon as ,Tex A ibu o 1.0
o e sacomp ehensi eapp oach,includingin e ac i es a is icalanalysiso  he
lexical,mo phological,syn ac ic,andseman ics uc u eo  he ex in eal- ime
ac oss15pa ame e s.
A ibu ion o  ex s is ac i ely de eloped in o eign ex ual s udies, in
pa icula , Bu ows’sme hod(Bu ows2002; A gamon2007;Ede ,Rybicki
2013; Bu ows e al. 2014) e ec i eness is es ed in many s udies on la ge
olumeso  ex ualda ao  a iouss yleslike:Englishp oseo  heea ly20 h
cen u y(Hoo e 2004);mode nEnglishpoe y(Hoo e 2005);poe icwo ks
inLa in(Rybicki,Ede 2011);p osewo kso majo gen esinEnglish,F ench,
I alian,Ge man,Polish,Hunga ian,aswellasLa inandA abic(Rybicki,Ede 
2011;E e e al.2015;Jannidise al.2015);poli ical ex sinEnglish,including

NATALIIA DARCHUK, OKSANA ZUBAN, VALENTYNA ROBEIKO,
YULIIA TSYHVINTSEVA, VICTOR SOROKIN, MYKOLA SAZHOK
228 Ac aLinguis icaLi huanicaXCI
hea ibu iono speecheso Ame icanp esiden s(Sa oy2015).Onemo e
au oma ed linguis ic analysis me hod called “Linguis ic Inqui y and Wo d
Coun ”(LIWC)isimplemen edasacomme cialapplica ionandadap ed o
se e allanguages(Meie e al.2019;LIWC-22).
In ecen yea s, he askso au oma ic ex a ibu ionha ep edominan ly
usedei he  ule-baseds ylome icme hodso deeplea ningme hods(Ede 
2015;Canhasie al.2022).Bo hme hodswe eimplemen edin heTex A ibu o 
sys em,e ec i ely unc ioningbu yieldingdi e en ou comes:1)The ule-
basedme hodallowsau oma ing hegene a iono linguis icconclusionsabou 
idiolec and ex  oxici y;2)Themachinelea ningme hodde e minesonly he
deg eeo  oxici yand hesimila i yo  ex sandcanbeusedinclassi ica ion
asks o moni o ing ex ualcon en .
The ecen  machine lea ning me hods o  sen imen  analysis and ex 
au ho shipiden i ica iona ebasedons a is icalapp oaches(TF-IDF,La en 
Di ichle Alloca ion)anddeeplea ning echniqueswi h a iousa chi ec u es
(CNN,LSTM,BERT)incombina ionwi hs ylome icme hods(Gup ae al.
2019;Canhasie al.2022;Bone ie al.2023).The epo ed esul sa eex emely
dependen on henumbe o classes( ypeso  oxici y,au ho s),gen e,leng h
o  he ex and,mos impo an ly, he olumeo  hep ope lyanno a ed ex 
co pusused o  he ainingp ocedu e.Thus,highaccu acy a esin he asks
o de ec ing oxici yandha espeechexceed90%usingaco puscon aining
enso  housandso documen s(Alkomah,Ma2022).Fo example,mo e han
90%accu acyinau ho shipa ibu ionisachie ed o English-languageli e a y
wo ksac oss1000au ho s,usingaco puso 6000documen s o 15au ho s.In
u n, o co po acon ainingless han1500documen s he epo edaccu acy
isabou 70%(Khane al.2023).Fo  heUk ainianlanguage, ex au ho ship
iden i ica ion and sen imen / oxici y analysis a e unde s udied p oblems
(Lupeie al.2020),mo eo e ,suchsys emsdono p oducelinguis icexpe ise
andcanno beusedas ools o linguis ic esea ch.
3. GENERALCHARACTERISTICSOFTHE
TEXTATTRIBUTORSYSTEMOPERATING
Du ing he de elopmen  o  Tex A ibu o , he ollowing asks we e
accomplished:1)Theo e icalLinguis ics:Ame hodologywasde eloped o 
o malizedmo phological,s a is ical,s ylome ic,a ibu ional,andsen imen 
analyses;2)So wa eEnginee ing:Thes uc u eo linguis icda abasesand
he so wa e o  he a o emen ioned au oma ic analyses we e de eloped;
3)Expe imen alLinguis ics:Thesys emwas es edonUk ainianmedia ex s,and
 S aipsniai / A icles 229
The Sys em o Au oma ic S ylome ic Analysis
o Uk ainian Media Tex s Tex A ibu o 1.0
(Techniques, Means, Func ionali y)
ananaly icalmodule o au oma ics ylome icexamina ionwasimplemen ed.
Theau oma edlinguis icsys em,implemen edasawebapplica ion,p ocesses
use -en e ed media-s yle ex s, gene a ing nume ical alues o  s a is ical
pa ame e s ha cha ac e ize he ex ’sa ibu es.Thesys em ollowsas uc u ed
ope a ionsequence:
1.Tex Tokeniza ion:B eakingdown he ex in osen encesandwo ds o 
analysis.
2.Mo phological Labeling: Assigning labels o wo ds based on hei 
g amma ical o ms.
3.Con ex ual Analysis: Upda ing he mo phological labels conce ning
con ex .
4.Syn ac ic Analysis:Es ablishingbina y ela ionshipsbe weenwo dsin
sen ences ounde s and hei g amma icals uc u e.
5.Syn ac ic Rela ion Es ablishmen :De e miningsyn ac icconnec ions
basedonp ede ined ules.
6.Emo ionally Nega i e Vocabula y Ma ching:Iden i yingwo ds oma
p ede inedse o nega i e ocabula y.
7.S a is ical Pa ame e E alua ion:Assessings a is icalpa ame e s o  he
inpu  ex usingau oma icallycompiled equency ocabula ies.
8.Visualiza ion o Resul s:P esen ings a is icalpa ame e iza ionou comes
g aphically,wi hempi ical aluesandcon idencein e als.
9.Compa ison o Resul s:Visualizings a is icalpa ame e iza ionou comes
o  womedia-s yle ex s, acili a ingcompa ison.
10.Euclidean Dis ance E alua ion:Quan i ying hedissimila i ybe ween
womedia-s yle ex s.
11.Expe Opinion Gene a ion:Gene a ing woexpe opinions:oneon
ex a ibu ionandano he on ex  oxici y h oughlinguis icexamina ion.
12.Toxici y De ec ion and Au ho ship Iden i ica ion:Le e agingmachine
lea ning echniques, pa icula ly neu al ne wo k models, o  de ec ing
oxici yandiden i yingau ho s.
The esul s o  he sys em a e o ganized in he ollowing pa i ions: Tex 
A ibu ion Index G oup, Tex  A ibu ion Expe  Opinion, Compa ison o 
Tex A ibu ion,Linguis icExpe iseo Tex Toxici y,andNeu alNe wo k
Opinions.
In Tex A ibu ion Index G oup (Fig. 1), s a is ical pa ame e s a e
o ganizedbasedon heinpu  ex :column1displays heIndex i le,column
2p esen s heempi icalnume ic alue,andcolumn3p o idesa isual e e ence
by compa ing he empi ical alue o  he index wi h con idence in e al
NATALIIA DARCHUK, OKSANA ZUBAN, VALENTYNA ROBEIKO,
YULIIA TSYHVINTSEVA, VICTOR SOROKIN, MYKOLA SAZHOK
230 Ac aLinguis icaLi huanicaXCI
h esholds(lowe anduppe )de i ed omUk ainianmedias yle ex s.The
empi icalnume ic alueis ep esen edon hescalebya illedin e ed iangle.
FIGURE 1:A agmen o  he ex a ibu ion esul
In Tex A ibu ion Expe Opinion (Fig. 2), we p esen  conclusions
ega ding he ypicalo indi iduallinguis ics a is ical ea u esex ac ed om
heanalyzed ex  o eaches ima edindex.Theseconclusionsa ede i ed om
answe s o he ques ion: Does henume ical alue o  he index allwi hin
hecon idencein e alo  hemedias yleo  heUk ainianlanguage?Based
onacompa isono nume ical alueswi h h eshold alueso  hecon idence
in e al, hesys emgene a es h eepossibleanswe s( ypicalsignso media
s yle,signso idios ylelowe  hanno malmedias yle,andsignso idios yle
highe  hanno malmedias yle).
In Compa ison o Tex A ibu ion (Fig. 4, subsec ion 4.2), upon
selec ing“Calcula eVec o Dis ance”, abula edda aisp omp lyupda edas
ollows:Column1lis suse -en e ed ex salongwi h hep e iouslyanalyzed
ex ; Column 2 displays he numbe  o  sen ences in each ex ; Column
3shows hewo dcoun ;Column4p esen s heEuclideandis ancebe ween
heanalyzed ex andeach ex in he able;Column5,labelled“Compa e”,
ini ia esanau oma iccompa isono s a is icalindicesbe ween heanalyzed ex 
and heselec ed ex .Uponac i a ing heco esponding“Compa e”elemen ,
comp ehensi ein o ma ionon hes a is icalcompa isonbe ween wo ex sis
p o idedin he“Tex A ibu ionIndices”g oup(Fig.3).
 S aipsniai / A icles 231
The Sys em o Au oma ic S ylome ic Analysis
o Uk ainian Media Tex s Tex A ibu o 1.0
(Techniques, Means, Func ionali y)
FIGURE 2: Anexampleo gene a edexpe opinion
In Linguis ic Expe ise o Tex Toxici y (subsec ion 4.3), a lexical-
seman icin e p e a iono  hecalcula ed oxici yindexisp esen ed(Fig.6).
InNeu al Ne wo k Opinionswep esen  heou pu o ou deeplea ning
model,whichquan i ies ex  oxici yonascale om0(non- oxic) o1(high
oxici y).Wealsoassess ex simila i y oknownau ho s,wi h alues anging
om0(nosimila i y) o1(highsimila i y),basedon heau ho s hemodelwas
ainedon.Resul sa eshownonly o  es swi hasimila i ymeasu eexceeding
0.1.Ou cu en sys emope a esindemomode,u ilizingalimi edco puso 
au ho  ex s ogaugesimila i ywi huse -en e ed ex .
4. ATTRIBUTIONOFUKRAINIAN-
LANGUAGETEXTS:EXPERIMENTS,
RESULTSANDDISCUSSIONS
4.1. S a is icalPa ame e iza ionIndices
Thede elopeds ylome icmodelsol es wo asks:
1)Iden i yingindi idual ea u eso anau ho ’ss yle(s ylome y);
2)Assessing hesimila i ybe ween wo ex sbasedonlinguis icands a is ical
pa ame e s(a ibu ion).
NATALIIA DARCHUK, OKSANA ZUBAN, VALENTYNA ROBEIKO,
YULIIA TSYHVINTSEVA, VICTOR SOROKIN, MYKOLA SAZHOK
238 Ac aLinguis icaLi huanicaXCI
Today, he in o ma ion componen  has become pa icula ly impo an  in
hyb idwa a e.The e o e, hema e ialo ou s udywasa esea chco puso 
onlinemedia ex so poli icaldiscou sewi ha olumeo 10millionwo ds.
Weuse he e m“ oxic ex ”inab oadsense.These ex sa echa ac e izedby
ha assmen , h ea s,obsceni y,cybe -bullying, olling,andiden i y-basedha e
ex s,andcon ainemo iogens,whicha ephenomenaandobjec s ha cause
nega i eemo ionsinape son(e.g.wa ,ai aid,co up ion).
Thisp ojec aimed ode e mine hepo en ial o  he ealiza iono nega i e
sen imen inUk ainian ex sand ode elopanau oma edsys em o iden i ying
oxiccon en .This askcomp ised woconsecu i esub asks: i s ,de eloping
asys emo  oxic ex linguis icexamina ionwi hde e mina iono  he oxici y
indices h oughdic iona yanalysisand ule-basedme hods;second,cons uc ing
a aining da ase  o  machine lea ning pu poses and he de elopmen  o  a
neu alne wo k o p edic ing ex  oxici yindices.
Le uslooka  hecomple iono  he i s  ask.Thisin ol ed hecons uc ion
o alexicog aphicda abasehousing h eedis inc dic iona ies:
1)Emo iogendic iona y:Acompila iono 5000wo ds(acco ding o he
meaningso lexical-seman ic a ian s)wi hnega i esen imen  ones, a edon
ascaleo –2.Examplesincludeimmo al,impuni y,b ibe yands eal;
2) Ha e Speech Dic iona y: Comp ising 3000 wo ds, including names o 
indi iduals(1620),obscene e ms(613),andabusi elanguage(787).Examples
includewes e ne (западенець: ade oga o yname o people om hewes e n
egionso Uk aine)andhucks e ;
3) Toxic Ph ases: A collec ion o  1500 idioma ic exp essions con eying
nega i eemo ions.Examplesincludeg imaces like a monkey,kisses his a seand
opens his mou h.
Eachen yin heselis swasanno a edwi hseman ic ea u es,wi h heHa e
SpeechDic iona yha ing18 ea u esand heToxicPh aseslis ha ing26.Fo 
ins ance,wo dsin heHa eSpeechDic iona ywe e aggedwi hcha ac e is ics
like‘s’ o sexism,‘ ’ o  acism,and‘e’ o ageism.Thislexicog aphicda abase,
amul ipa ame icsys em,assignsseman ic ea u es ouni s(wo dso ph ases)
which,combinedwi h equencyda a omanalyzed ex ,allows o  oxici y
analysis.
The oxici yindexo  he ex iscompu edusing he o mula:
I ox = (e + |K| (m + )) / n * 10,
He e,nis he ex  olume;eis hecoun o emo ionalwo ds;mis hecoun 
o ha espeechwo ds; is hecoun o  oxicph ases;Kisacoe icien equal
o–2,i emphasisesha espeechwo dsand oxicph ases,whichonascaleo 

 S aipsniai / A icles 239
The Sys em o Au oma ic S ylome ic Analysis
o Uk ainian Media Tex s Tex A ibu o 1.0
(Techniques, Means, Func ionali y)
i edigi s(+2,+1,0,–1,–2)co esponds o–2.Index alues a y om0 o
+1.
Fo ins ance, he oxici yindexo  he ex “Peoplewi ha edpen(Людиз
червоноюручкою)”(12sen ences,129wo ds)is1.01,indica ingi shigh oxici y,
as he h eshold o suchsho  ex s anges om0.1 o0.7.Simul aneously,
hesys emgene a es he esul so anau oma iclinguis icexamina iono  he
ex  oxici y( e e  oFig.6),comp ising:1)as a is icalmapo  heseman ic
classeso  he nega i e ocabula yacco ding o heclassi ica ionma ke s o 
lexicog aphiclis s(emo iogens–5, ulga isms–2,sexism–2);2)a ex wi h
speci icwo dso nega i esen imen , e balizing heca ego ieso  hes a is ical
map.Le us akealooka anexce p o  heanalyzed eal-li e ex (Fig.6): hose
people who walk a ound wi h a ed pen and co ec mis akes in o he people’s
pos s: who e en a e you? Do you ha e any idea how annoying you a e? I s udy you
p o iles on pu pose, I’m cu ious abou he wo ld you exis in. This wo ld sca es me
a lo . I since ely hope ha in eal li e we will no c oss pa hs unde any ci cums ances.
(оці люди, які ходять з червоною ручкою і виправляють помилки в чужих
постах: ви хто взагалі такі? Ви хоч уявляеєте, як ви бісите? Я спеціально
вивчаю ваші профілі, мені цікаво в якому світі ви існуєте. Мене цей світ дуже
лякає. Щиро сподіваюся, що в реальному житті ми з вами не пересічемося
ні за яких обставин). He e, hesys emau oma icallyhighligh sinbold he
ollowingwo dswi hanega i ecomponen :mis akes (помилки), o he people’s
(чужих), annoying (бісите), sca es (лякає).
FIGURE6: A agmen o  ex  oxici yau oma edlinguis icexamina ion
NATALIIA DARCHUK, OKSANA ZUBAN, VALENTYNA ROBEIKO,
YULIIA TSYHVINTSEVA, VICTOR SOROKIN, MYKOLA SAZHOK
240 Ac aLinguis icaLi huanicaXCI
5. MACHINELEARNINGAPPROACHES:
EXPERIMENTS,RESULTSAND
DISCUSSIONS
Le us ocuson hesecondope a ionmodeo  heTex A ibu o sys em
dedica ed omachinelea ning echniques.A  heinpu  o model aining,we
ha e ex co po alabelledacco ding o hegoalo eachsub ask:(a) oxici y
de ec ion ex in o ma ionand(b) au ho ship iden i ica ion. Eachco pusis
di idedin o ainingandcon olse s,aswellas, o su icien lyla geco pus,
a alida ion se . The pa ame e s o  he chosen p oblem-sol ing model a e
es ima edon he ainingse .Hype pa ame e so  hemodela eadjus edby he
alida ionse .The inalpe o manceindica o so  hemodela emeasu edona
con olse , akingin oaccoun  hea ailabili yo compu ing esou ces,which
a enecessa y o  hemodel’sope abili y.Aco pus o eachsub askhasbeen
de elopedinpa allelwi h heTex A ibu o machinelea ningcomponen ,so
hecu en modelswe e ainedon ela i elysmallda aincludingsho In e ne 
ex so Uk ainian-languageblogs,commen s,a icles,e c.
5.1. Toxici yde ec ion
Asa esul o ase ieso expe imen als udies,acompu a ionallye icien 
a chi ec u ewaschosenbasedon he as Tex me hodandi s ools(Jouline al.
2016).Thisme hodp o ideswo dembeddingsandes ima es hep obabili y
dis ibu iono documen sacco ding op ede inedclasses.Wo dsa ep esen ed
in ec o  o mbasedonau oma icwo dspli ingin opa s(subwo ds),which
simula es heopennesso  hedic iona y.Thesubwo dp esen a ionisimpo an 
since heUk ainianlanguageishighlyin lec i e,andplen yo unseenwo dsas
wellaswo dswi hspellinge o smus beco e ed.Thee ec i enesso  heused
app oachisalsode e minedby he echnicalcondi ionso sys emope a ionand
a ailable ex ual esou ces o model aining.
Ap epa edco puso app oxima ely12,000 ex documen swasused o bina y
classi ica ion ode ec  oxiccon en .Thebes  esul ,basedon hegene alized
me ic(F1=79.4%),wasachie edwi h he ollowinghype pa ame e se ings:
awo d ec o dimensionali yo 56,anini iallea ning a eo 0.15,500 aining
epochs,lexicalcon ex  ep esen edbybig ams,subwo dleng hs anging om
2 o 5 cha ac e s, and a decision h eshold o  0.4. Addi ionally, he model
demons a edasensi i i yo 85%whileachie inganaccu acyo app oxima ely
70%.
 S aipsniai / A icles 241
The Sys em o Au oma ic S ylome ic Analysis
o Uk ainian Media Tex s Tex A ibu o 1.0
(Techniques, Means, Func ionali y)
5.2. Au ho shipiden i ica ion
Expe imen al esea ch o  models o  ex  au ho ship iden i ica ion was
ca iedou usingapa o  heco puso publiclya ailablesocio-poli ical ex s,
inwhichi ispossible oiden i y hepe sonwhois heau ho o  hedocumen .
A o alo 702 ex s om ou au ho swhopublished hela ges numbe o 
documen sbe ween60and1000cha ac e swe eselec ed.Thela ges numbe 
o publica ionsby heau ho is304(115 housandcha ac e s),and hesmalles 
is109(49 housandcha ac e s).Fo  es ing,25 ex s o eachselec edau ho 
we e andomlysampled.The es o  hedocumen smadeup he ainingse .
Thebes  esul ,acco ding o hegene alizedme ic(F1=81%),wasachie ed
wi h he ollowinghype pa ame e  alues:awo dembeddingspacedimension
o  50, an ini ial lea ning a e o  0.1, 50 aining epochs, lexical con ex 
ep esen edbybig ams,subwo dleng hs anging om2 o5cha ac e s,and
adecision-making h esholdo 0.5.No ably, o ce ainmodelcon igu a ions,
hesensi i i y o heau ho wi h hela ges numbe o documen sin heda ase 
eached100%,wi hanaccu acyexceeding90%.These esul sdemons a e he
easibili yo de elopingane ec i esys emcapableo accu a elyco e ing ex s
byau ho swhoa esu icien ly ep esen edin he ainingse .
The ainedneu alne wo kmodelswe ein eg a edin o heTex A ibu o 
sys emin he“clien -se e ”a chi ec u e,whe e he esul so  wo asksa e
p esen ed:(a) oxici yde e mina ionand(b)au ho shipde ec ion.The esul 
o  hesys em’swo kon he ex o  he“Lawo Uk aineonHighe Educa ion”:
Toxici yindex(es ima ein he o mo p obabili y: how oxic he ex is)–
0.04, simila i y wi h au ho s (es ima ions in he o m o  he p obabili y o 
au ho ship o au ho sknown o hesys em):0.49 o I.Fa ion,0.30 o Oleksii
Honcha enko,and0.22 o Ma ianaBezuhla.
6. CONCLUSIONS
TheTex A ibu o sys emiscu en lyunde going es ingphases oadd ess
a ibu ionand oxici yde e mina ionissueswi hinUk ainian-language ex s.
Thissys emisbo hcon enien ande ec i e o  a ious echnologicals udies.
O e 5mon hs, heTex A ibu o webapplica ionhasau oma icallyanalyzed
784Uk ainian ex sby226use s.
Ou  indingsdemons a e hee ec i enesso ou me hod,whichin ol es
quan izing e balelemen sbasedon o malg amma icalandseman icpa ame e s,
pa icula ly o nega i eemo ionali y.Thisisachie ed h oughacombina ion
o dic iona yand ule-basedapp oaches,complemen edbymachinelea ning
NATALIIA DARCHUK, OKSANA ZUBAN, VALENTYNA ROBEIKO,
YULIIA TSYHVINTSEVA, VICTOR SOROKIN, MYKOLA SAZHOK
242 Ac aLinguis icaLi huanicaXCI
echniques.Expe imen almachinelea ning esea ch o  oxici yde ec ionand
au ho ship iden i ica ion by ex  allowed us oob ain esul s compa able o
esul so expe imen swi hsimila modelcha ac e is ics(numbe o documen s
andau ho s) epo ed o o he languages.Subwo dmodelsshowedenhanced
obus nessagains lexicalopennessand ex uale o s.
Theconduc ed esea chpa es heway o sol ingsuchp oblems o Uk ainian
as de ec ingand moni o ing oxic con en ,ha e speechandmisin o ma ion,
au ho ship e i ica ion and deob usca ion, au ho  p o iling and dia izing,
psycholinguis ic p o iling, s yle modelling, and acking he sou ce o  ake
con en ,e c.Thede eloped ex analysisanda ibu ionschemecanalsobe
applied oo he languages.
Lookingahead,weplan oin eg a eou au oma ic ex a ibu ionand oxici y
de ec ion ool wi h seman ic, syn ac ic, psycholinguis ic, and sociolinguis ic
analysis.Thisin eg a ionwillgi eusadeepe unde s andingo  heimpac on
he eade .Byusingseman icanalysiso lexical axonomy,weaim oiden i y
na a i e elemen s inhe en  in he ex , he eby inc easing he eliabili y o 
ex a ibu ioninau ho shipiden i ica ionp ocedu es.Inaddi ion,weplan
oexpandou  ex co po aandu ilizemul ilingualla gelanguagemodels o
u he ex end hecapabili ieso ou sys emaswellas oimplemen  ools o 
s a is ical ex compa ison o alls yleso  heUk ainianlanguage.
Acknowledgemen s
The sys em has been c ea ed wi hin a p ojec  suppo ed by he B i ish
EmbassyinKyi andca iedou bya eamo educa o sand esea che so  he
Uk ainianLanguageandAppliedLinguis icsChai ,Ta asShe chenkoNa ional
Uni e si yo Kyi .
Wewouldlike oexp essou since eapp ecia ion o heB i ishEmbassy
inKyi  o  hei  inancialsuppo o  his esea chp ojec .Wea eg a e ul o
S i lanaYa o skaandI ynaBezko o ayna o  hei guidanceandassis ance
h oughou  he p ojec . Special hanks o Kos ian yn Goncha enko o  he
o ganiza ion and documen a y suppo  o  he p ojec . Fu he mo e, we a e
g a e ul o he s uden s o  he educa ional p og am “Applied (Compu e )
Linguis icsandEnglishLanguage”o Ta asShe chenkoNa ionalUni e si y
o Kyi .Theyha econ ibu ed hei  ime,expe ise,ande o s o o ming
heco puso Uk ainian oxic ex s.Wea eg a e ul oOksana Tolochko,a
g adua eo ou bachelo ’sdeg eep og am, o c ea inga onaldic iona yo  he
Uk ainianlanguage,whichweused ocompilealexicog aphiclis o nega i e
emo iogens.Addi ionally,wea edeeplyapp ecia i eo  hee o smadeby
MykolaKos iko in heda acollec ionp ocess.Wewouldlike oexp essou 
since eg a i ude oallmembe so  he esea chcommuni ywhocon ibu ed o
hiss udywi h hei ad iceandconsul a ions.
 S aipsniai / A icles 243
The Sys em o Au oma ic S ylome ic Analysis
o Uk ainian Media Tex s Tex A ibu o 1.0
(Techniques, Means, Func ionali y)
REFERENCES
ALIAS–Au oma ed Linguis ic Iden i ica ion & Assessmen Sys em.A ailablea :h ps://
alias echnology.com/alias-o e iew/.
Alkomah Fa imah, Ma Xiaogang 2022:ALi e a u eRe iewo Tex ualHa e
Speech De ec ion Me hods and Da ase s. – In o ma ion 13(6), 273. DOI: 10.3390/
in o13060273.
A gamonShlomo2007:In e p e ingBu ows’sDel a:Geome icandP obabilis ic
Founda ions. – Li e a y and Linguis ic Compu ing 23, 131–147. DOI: 10.1093/llc/
qn003.
Bone i And ea, Ma ínez-Sobe  Ma celino, To es Julio, Vega Jose,
Pelle in Sebas ien, Vila-F ancés Joan 2023:Compa isonbe weenMachine
Lea ningandDeepLea ningApp oaches o  heDe ec iono ToxicCommen son
SocialNe wo ks.–Applied Sciences13(10),6038.DOI:10.3390/app13106038.
Buk Solomija 2021: Long p ose ic ion by I. F anko: elec onic co pus, equency
dic iona ies and o he in e disciplina y con ex s,LNUimeniI anaF anka[I anF anko
Na ionalUni e si yo L i ].
Bu ows John 2002: “Del a”: a Measu e o  S ylis ic Di e ence and a Guide o
LikelyAu ho ship.–Li e a y and Linguis ic Compu ing17(3),267–287.DOI:10.1093/
llc/17.3.267.
Bu ows S e en, Ui denboge d Alexand a, Tu pin And ew 2014:
Compa ing echniques o au ho shipa ibu iono sou cecode.–So wa e: P ac ice
and Expe ience44(1),1–32.DOI:10.1002/spe.2146.
CanhasiE can,Kad iuA bana,MisiniA a2022:ASu eyonAu ho ship
Analysis Tasks and Techniques. – SEEU Re iew 17(2), 153–167. DOI: 10.2478/
seeu -2022-0100.
Chuhuno Vadym2009:Diahnos yka psykho e apii a psykho e ape ychnyi diahnoz,
Kha ki :Nauka.
Da chuk Na alia, So okin Vik o  2022: Pa ame e iza ion o  he Uk ainian
Tex  Co pus based on pa sing. – Compu a ional Linguis ics and In elligen Sys ems,
P oceedings o  he 6 h In e na ional Con e ence on Compu a ional Linguis ics and
In elligen Sys ems(COLINS-2022)1:Main Con e ence,eds.V.Ly yn,N.Sha ono a,
I.Jonek-Kowalska,A.Kowalska-S yczen,V.Vyso ska,Ye.Kup iiano ,O.Kanishche a,
O.Che ednichenko,Th.Hamon,N.G aba ,Poland,12–13May,2022,256–265.
Da chuk Na alia, Zuban’ Oksana, So okin Vik o  2024: On s ylis ic
di e en ia ion in Uk ainian poe ic speech based on he syn ac ic s uc u e o 

NATALIIA DARCHUK, OKSANA ZUBAN, VALENTYNA ROBEIKO,
YULIIA TSYHVINTSEVA, VICTOR SOROKIN, MYKOLA SAZHOK
244 Ac aLinguis icaLi huanicaXCI
sen ences.–Bulle in o Ta as She chenko Na ional Uni e si y o Kyi . Li e a y S udies.
Linguis ics. Folklo e S udies1(35),5–10.DOI:10.17721/1728-2659.2024.35.01.
Da čuk Na alija 2013: Kompju e ne ano u annja uk ajins’koho eks u: ezul’ a y i
pe spek y y,Kyi :Os i aUk ajiny.
Da čukNa alija,Vasyl’je aI yna,Vasyl’je Oleksij2021:Vek o namodel’
analizus ylis yky eks i .–Uk ajins’kyj izyčnyj žu nal66(5),373–378.
Ede  Maciej 2015:Doessizema e ?Au ho shipa ibu ion,smallsamples,big
p oblem.–Digi al Schola ship in he Humani ies30(2),167–182.DOI:10.1093/llc/
q 066.
Ede  Maciej, Rybicki Jan 2013:Dobi dso a ea he  eally lock oge he ,o 
how ochoose ainingsamples o au ho ship a ibu ion.–Li e a y and Linguis ic
Compu ing28(2),229–236.DOI:10.1093/llc/ qs036.
E e S e an,P oislThomas,SchöchCh is o ,JannidisFo is,Piels öm
S e en, Vi  Tho s en 2015:ExplainingDel a,o :Howdodis ancemeasu es
o  au ho ship a ibu ion wo k? – Co pus Linguis ics 2015: Abs ac  Book, Uni ed
Kingdom,21July2015,eds.F.Fo ma o,A.Ha die,Lancas e :UCREL.A ailablea :
h ps://zenodo.o g/ eco ds/18308.
Gup aSh iyaT.P.,SahooJaja iK.,RoulRajend aK.2019:Au ho ship
iden i ica ionusing ecu en neu alne wo ks.–ICISDM’19:P oceedingso  he2019
3 dIn e na ionalCon e enceonIn o ma ionSys emandDa aMining,Uni edS a es,
06 Ap il 2019, New Yo k: Associa ion o  Compu ing Machine y, 133–137. DOI:
10.1145/3325917.3325935.
Hoo e  Da id 2004:Tes ingBu ows’sdel a.–Li e a y and Linguis ic Compu ing
19(4),453–475.DOI:10.1093/llc/19.4.453.
Hoo e Da id2005:Del a,Del aP ime,andMode nAme icanPoe y:Au ho ship
A ibu ionTheo yandMe hod.–ACH/ALLC 2005:P oceedingso  he17 hJoin 
In e na ionalCon e enceo  heAssocia ion o Compu e sand heHumani ies(ACH)
and heAssocia ion o Li e a yandLinguis icCompu ing,Canada,15–18June2005,
Uni e si yo Vic o ia:Humani iesCompu ingandMediaCen e,79–80.
JannidisFo is,Piels ömS e en,SchöchCh is o ,Vi Tho s en2015:
Imp o ingBu ows’Del a–Anempi icale alua iono  ex dis ancemeasu es.–DH
2015:Digi alHumani ies,Abs ac so  heGlobalDigi alHumani iesCon e ence11,
Aus alia,29June–3July2015,Sydney.DOI:h ps://zenodo.o g/ eco ds/1321296.
Joulin A mand, G a e Edoua d, Bojanowski Pio , Douze Ma hijs,
JégouHé e,Mikolo Tomas2016:Fas Tex .zip:Comp essing ex classi ica ion
models.–A Xi e-p in s 1612.03651.DOI:10.48550/a Xi .1612.03651.
 S aipsniai / A icles 245
The Sys em o Au oma ic S ylome ic Analysis
o Uk ainian Media Tex s Tex A ibu o 1.0
(Techniques, Means, Func ionali y)
Ka aso  Vi alii, Le chenko Olena 2024:S a is icalp o ileo O.Zabuzhko’s
idios yle. – CEUR Wo kshop P oceedings 3722: P oceedings o he 8 h In e na ional
Con e ence on Compu a ional Linguis ics and In elligen Sys ems,L i ,Ap il12–13,2024,
251–264.
KhanTalhaF.,Anwa Waheed,A shadHume a,AbbasSyedN.2023:
An Empi ical S udy on Au ho ship Ve i ica ion o  Low Resou ce Language Using
Hype -Tuned CNN App oach. – IEEE Access 11, 80403–80415. DOI: 10.1109/
ACCESS.2023.3299565.
Khomy skaI yna,TeslyukVasyl,Bazyle ychI yna,Ka amyshe aI yna
2023:Au oma edIden i ica iono Au ho ialS yles.–CEUR Wo kshop P oceedings:
P oceedings o he 7 h In e na ional Con e ence on Compu a ional Linguis ics and In elligen
Sys ems, 3396, Kha ki , Ap il 20–21, 2023. A ailable a : h ps://ceu -ws.o g/Vol-
3396/pape 26.pd .
KUM–Ko pus uk ajins’koji mo y:Linh is yčnyjpo alMo a.in o.A ailablea :h p://
www.mo a.in o/co pus.aspx.
LIWC-22–Linguis ic Inqui y and Wo d Coun .A ailablea :h ps://www.liwc.app.
Lo o ska Na aliia 2022:S a is icalCha ac e is icso RomanI anychuk’sIdiolec 
(BasedonW i e ’sTex Co pus).–CEUR Wo kshop P oceedings3171,487–500.
LupeiMaksym,Mi saAlexande ,Repa iukVolodymy ,Sha kanVasyl
2020: Iden i ica ion o  au ho ship o  Uk ainian-language ex s o  jou nalis ic s yle
usingneu alne wo ks.–Eas e n-Eu opean Jou nal o En e p ise Technologies1/2(103),
30–36.DOI:10.15587/1729-4061.2020.195041.
McInnes Leland, Healy John, Mel ille James 2020: UMAP: Uni o m
Mani oldApp oxima ionandP ojec ion o DimensionReduc ion.–A Xi e-p in s
1802.03426.DOI:10.48550/a Xi .1802.03426.
Meie Tabea,BoydRyan,Pennebake James,MehMa hiasl,Ma in
Mike,Wol Ma kus,Ho nAnd ea2019:“LIWCau Deu sch”:Thede elopmen ,
psychome ics, and in oduc ion o  DE-LIWC2015. – PsyA Xi P ep in s. DOI:
10.31234/os .io/uq8z .
Rybicki Jan, Ede  Maciej 2011: Deepe  Del a ac oss Gen es and Languages:
DoWeReallyNeed heMos F equen Wo ds?–Li e a y and Linguis ic Compu ing
26(3),315–321.DOI:10.1093/llc/ q 031.
Sa oy Jacques 2015: Compa a i e e alua ion o  e m selec ion unc ions o 
au ho shipa ibu ion.–Digi al Schola ship in he Humani ies30(2),246–261.DOI:
10.1093/llc/ q 047.
NATALIIA DARCHUK, OKSANA ZUBAN, VALENTYNA ROBEIKO,
YULIIA TSYHVINTSEVA, VICTOR SOROKIN, MYKOLA SAZHOK
246 Ac aLinguis icaLi huanicaXCI
Tex A ibu o  1.0 2024: Tex A ibu o 1.0: The sys em o au oma ic s ylome ic
analysis o Uk ainian media ex s.A ailablea : a.mo a.in o.
Volos Yana, Le chenko Olena 2023:S a is icalp o ileo Vale ieShe chuk’s
idios yle(mo phologicalle el).–Slobozhan Scien i ic Bulle in, Se . Philology,3,32–36.
DOI:10.32782/philspu/2023.3.6.
Zuban’Oksana2019:TheMo phemicSys emS ylome icAnalysiso  heUk ainian
Poe s’ Idios yles: Co pus Based App oach. – Linh is yčni s udiji 38, 96–104. DOI:
10.31558/1815-3070.2019.38.15.
Uk ainosžiniasklaidos eks ųau oma inės
s ilome inėsanalizėssis ema
„Tex A ibu o 1.0“
(me odai,p iemonės, unkcionalumas)
SANTRAUKA
Sis ema „Tex A ibu o “ s a is iškai pa ame izuoja uk ainiečių kalbos eks us au o-
ys ėsa pažinimoi nuo aikųanalizėsaspek ais.Naudodamadaugiapa ame inį15s a is i-
niųindeksų inkinį,„Tex A ibu o “apibūdinaau o iness ilis inesypa ybes.Jiin eg uoja
okiusme oduskaipkomponen inėanalizė,paski s ymoanalizė,k an a imasi nuo ai-
kųanalizėpanaudojan  žodynusi  mašininiomokymosime odus, ski usnemandagaus
eks oi au o ys ėsišaiškinimui.Sis emaapdo oja eks us aikydama okeniza imą,mo -
ologinįžymėjimą,kon eks inęi sin aksinęanalizęi emociškaineigiamožodynoa i iki-
mą,leidžian įišsamiai izualizuo ii ekspe iškaiį e in i eks op isky imąi neigiamą
u inį.Ekspe imen aipa i inasis emosgebėjimąnus a y iunikaliusau o iausb uožus
i į e in i eks opanašumą,ypačpoli iniodisku soi nemandagioskomunikacijosžinias-
klaidoskon eks eRusijosi Uk ainoska ome u.S aipsnyjepab ėžiamap ak inėi  eo i-
nė„Tex A ibu o “ eikšmė,demons uojan jose ek y umądidelėms eks oapim imsi 
iksliomsanali inėmsgalimybėms.Sėkmingasžodynais, aisyklėmispag įs ųi mašininio
mokymosime odų aikymaspab ėžiasis emos i umąi uni e salumą.Be a e sijai  e-
be ykdomų y imų ezul a aiy anuodugniaiišnag inė i,obūsimosjųk yp ysnus a omos
siekian išplės ikalbinįko pusąi  obulin imašininiomokymosimodelius,siekian page-
in i ikslumąi p i aikomumą.
 S aipsniai / A icles 247
The Sys em o Au oma ic S ylome ic Analysis
o Uk ainian Media Tex s Tex A ibu o 1.0
(Techniques, Means, Func ionali y)
Į eik a2024m.lapk ičio25d.
NATALIIADARCHUK
Ta as She chenko Na ional Uni e si y o Kyi
60 Volodymy ska S ee
Kyi , 01033, Uk aine
n.da [email protected]
OKSANAZUBAN
Ta as She chenko Na ional Uni e si y o Kyi
60 Volodymy ska S ee
Kyi , 01033, Uk aine
[email protected]
VALENTYNAROBEIKO
Ta as She chenko Na ional Uni e si y o Kyi
60 Volodymy ska S ee
Kyi , 01033, Uk aine
alen yna. obeik[email p o ec ed]
YULIIATSYHVINTSEVA
Ta as She chenko Na ional Uni e si y o Kyi
60 Volodymy ska S ee
Kyi , 01033, Uk aine
Ins i u e o he Uk ainian Language o
he Na ional Academy o Sciences o Uk aine
4 Mykhailo H ushe skyi S ee
Kyi , 01001, Uk aine
juli o[email p o ec ed]
VICTORSOROKIN
Ta as She chenko Na ional Uni e si y o Kyi
60 Volodymy ska S ee
Kyi , 01033, Uk aine
ic o .so ok[email p o ec ed]
MYKOLASAZHOK
Ins i u e o in o ma ion echnologies and sys ems
o he Na ional Academy o Sciences o Uk aine
40 Akademika Hlushko a A enue
Kyi , 03187, Uk aine
s[email p o ec ed]om