Ukraine War: an online corpus to analyze the impact of the war in Ukraine

Uni e sidade do Minho Escola de Engenha ia Depa amen o de In o má ica Ana Ri a Mi anda Rosendo Uk aine Wa : An Online Co pus o analyze he impac o he Wa in Uk aine Feb ua y 2024 Uni e sidade do Minho Escola de Engenha ia Depa amen o de In o má ica Ana Ri a Mi anda Rosendo Uk aine Wa : An Online Co pus o analyze he impac o he Wa in Uk aine Mas e disse a ion In eg a ed Mas e ’s in In o ma ics Enginee ing Disse a ion supe ised by Ped o Rangel Hen iques C is iana A aújo Feb ua y 2024 i AUTHOR COPYRIGHTS AND TERMS OF USAGE BY THIRD PARTIES This is an academic wo k which can be u ilized by hi d pa ies gi en ha he ules and good p ac ices in e na ionally accep ed, ega ding au ho copy igh s and ela ed copy igh s. The e o e, he p esen wo k can be u ilized acco ding o he e ms p o ided in he license bellow. I he use needs pe mission o use he wo k in condi ions no o eseen by he licensing indica ed, he use should con ac he au ho , h ough he Reposi ó iUM o Uni e si y o Minho. License p o ided o he use s o his wo k A ibu ion-NonComme cial CC BY-NC h ps://c ea i ecommons.o g/licenses/by-nc/4.0/ ii STATEMENT OF INTEGRITY I he eby decla e ha ing conduc ed his academic wo k wi h in eg i y. I con i m ha I ha e no used plagia ism o any o m o undue use o in o ma ion o alsi ica ion o esul s along he p ocess leading o i s elabo a ion. I u he decla e ha I ha e ully acknowledged he Code o E hical Conduc o he Uni e si y o Minho. Ana Ri a Mi anda Rosendo ACKNOWLEDGMENTS I would like o exp ess my g a i ude o all o hose who made subs an ial con ibu ions o he success ul conclusion o his wo k. Fi s ly, I would like o hank he ad iso , P o esso Ped o Rangel Hen iques, and he supe iso , C is iana A aújo, o he oppo uni y. I would also like o hank bo h o hem o hei guidance, ad ice, and help, as well as hei con inuous suppo h oughou his p ocess. Thei guidance was undamen al o he de elopmen and success o his wo k. To my amily, I exp ess my deep g a i ude o hei cons an suppo and encou agemen h oughou his p ocess. Finally, I exp ess my g a i ude o all my iends o hei in aluable suppo h oughou his jou ney, be i academically o socially. They p o ided s ess- ee momen s and enjoyable imes ha elie ed he p essu es o he wo k. iii ABSTRACT This documen epo s a Mas e ’s wo k, he inal p ojec o he 5 h yea o he In eg a ed Mas e ’s in In o ma ics Enginee ing, ha was accomplished a Uni e sidade do Minho in B aga, Po ugal. On Feb ua y 24,2022, a con lic be ween wo coun ies, Uk aine and Russia, began. The wa be ween wo coun ies is de as a ing and a ec s many people, bo h esiden s o he coun ies di ec ly in ol ed and neighbo ing coun ies. As a highly signi ican e en , i ga he s co e age om many sou ces globally, including adi ional p in newspape s, online news pla o ms, social ne wo ks, blogs, ele ision p og ams, and mo e. Howe e , all o his in o ma ion is sca e ed ac oss di e en websi es and social ne wo ks. I esea che s (in he a eas o Linguis ics, His o y, Humani ies, e c.) and cu ious people wan o analyze his da a, hei wo k will be e y di icul . The e o e, i is essen ial o ga he he in o ma ion on a single pla o m. This wo k aims o c ea e an online co pus in he Po uguese language ega ding he Uk aine Wa , based on Po uguese online newspape s’ news as well as commen s on social media. To ul ill he goal o his wo k, ini ially, a a ie y o news sou ces we e conside ed, and he Po uguese online newspape s “Público” and “Jo nal de Negócios” we e selec ed, as well as he pla o m “Reddi ”. To ex ac he equi ed in o ma ion, he echnique o Web Sc aping was used. The e o e, o each sou ce, an ex ac o was de eloped ha ex ac ed he necessa y in o ma ion and sa ed i in a JSON ile. Following ha , Na u al Language P ocessing Techniques we e used o p ocess he ga he ed in o ma ion. A e wa d, he ex ac ed in o ma ion was s o ed in a non- ela ional da abase, MongoDB. Finally, a websi e called GUCO was designed and implemen ed, p o iding use s wi h he capabili y o na iga e and explo e he c ea ed co pus. The GUCO websi e is a ailable a he add ess: h ps://guco.epl.di.uminho.p /. Keywo ds: Online Co pus, Uk aine Wa , Rebuild he Wa h ough he News, Social Ne wo k Analysis, Web Sc aping, Na u al Language P ocessing i RESUMO Es e documen o ela a um abalho de mes ado, o p oje o inal do 5º ano do Mes ado In eg ado em Engenha ia In o má ica, que se á ealizado na Uni e sidade do Minho em B aga, Po ugal. Em 24 de e e ei o de 2022, o con lic o en e dois países, Uc ânia e Rússia, começou. A gue a en e dois países é de as ado a e a ec a mui as pessoas, an o esiden es nos países di ec amen e en ol idos como nos países izinhos. Sendo um e en o al amen e impo an e, es e eúne cobe u a de mui as on es po odo o mundo, incluindo jo nais adicionais imp essos, pla a o mas de no ícias online, edes sociais, blogs, p og amas de ele isão e mui o mais. No en an o, oda es a in o mação es á espalhada po di e en es websi es e edes sociais. Se pesquisado es (das á eas de Linguís ica, His ó ia, Ciências Humanas, e c.) e cu iosos quise em analisa esses dados, o seu abalho se á mui o di ícil. Po an o, é undamen al euni as in o mações numa única pla a o ma. Es e abalho em como obje i o a c iação de um co pus online na língua po uguesa sob e a Gue a da Uc ânia baseado em no ícias de jo nais online po ugueses, bem como em comen á ios nas edes sociais. Pa a cump i o obje i o des e abalho, inicialmen e, o am conside adas di e sas on es de no ícias, e os jo nais online po ugueses “Público” e “Jo nal de Negócios” o am se- lecionados como on es de in o mação pa a o co pus, bem como pos s e comen á ios da pla a o ma “Reddi ”. Pa a ex ai a in o mação necessá ia, oi u ilizada a écnica de Web Sc aping. Po an o, pa a cada on e oi desen ol ido um ex a o que ex aiu as in o mações necessá ias e gua dou-as num ichei o JSON. De seguida, o am u ilizadas Técnicas de P o- cessamen o de Linguagem Na u al pa a p ocessa a in o mação ecolhida. Pos e io men e, os dados o am a mazenadas numa base de dados não elacional, MongoDB. Po im, oi desen ol ido um websi e pa a que o u ilizado possa explo a o co pus c iado. O websi e GUCO pode se acedido em: h ps://guco.epl.di.uminho.p /. Pala as-cha e: Co pus Online, Gue a da Uc ânia, Recons ução da Gue a a a és das No ícias, Análise de Redes Sociais, Web Sc aping, P ocessamen o de Linguagem Na u al CONTENTS 1In oduc ion 1 1.1Con ex 1 1.2Objec i es 2 1.3Me hodology App oach 3 1.4Resea ch Hypo hesis 3 1.5Documen S uc u e 3 2S a e o he A 5 2.1Online Co pus 5 2.2Compu e -Media ed Communica ion 5 2.3Web Sc aping 6 2.4Na u al Language P ocessing 7 2.5Rela ed p ojec s 8 3P oposed App oach 11 3.1Sys em A chi ec u e 11 3.2Ex ac o s 12 4GUCO, Co pus Cons uc ion 21 4.1Ex ac o s 21 4.1.1Público 21 4.1.2Jo nal de Negócios 25 4.1.3Reddi 29 4.2Gene al JSON Fo ma 32 4.3Da a P ocessing 35 4.3.1Público 35 4.3.2Jo nal de Negócios 36 4.3.3Reddi 37 5GUCO, Pla o m 39 5.1Mockups 39 5.2Backend 43 5.3F on end 45 5.4A isi o he web pla o m 47 6Conclusion 55 6.1Fu u e Wo k 56 i LIST OF FIGURES Figu e 1Sys em A chi ec u e 12 Figu e 2Example o a news publica ion as an inpu o he ex ac o 14 Figu e 3 Example o he commen s o a news publica ion as an inpu o he ex ac o 15 Figu e 4(Mockup) GUCO Websi e: Home Page 40 Figu e 5(Mockup) GUCO Websi e Menu: P ojec Page Op ion 40 Figu e 6(Mockup) GUCO Websi e Menu: Ex ac o s Page Op ion 41 Figu e 7(Mockup) GUCO Websi e Co pus: Co pus Page 41 Figu e 8(Mockup) GUCO Websi e Co pus: Lis o esul s 42 Figu e 9(Mockup) GUCO Websi e Co pus: Resul Page 43 Figu e 10 GUCO Websi e: Home Page 48 Figu e 11 GUCO Websi e: In oduc ion Page 48 Figu e 12 GUCO Websi e: Aims Page 49 Figu e 13 GUCO Websi e: Sys em A chi ec u e Page 49 Figu e 14 GUCO Websi e: “Público” Ex ac o Page 50 Figu e 15 GUCO Websi e: “Jo nal de Negócios” Ex ac o Page 50 Figu e 16 GUCO Websi e: “Reddi ” Ex ac o Page 51 Figu e 17 GUCO Websi e: Publica ions Page 51 Figu e 18 GUCO Websi e: Team Page 52 Figu e 19 GUCO Websi e Co pus: Co pus Page 53 Figu e 20 GUCO Websi e Co pus: Lis o Resul s Page 53 Figu e 21 GUCO Websi e Co pus: Resul Page (1)54 Figu e 22 GUCO Websi e Co pus: Resul Page (2)54 ii 2 STATE OF THE ART In his sec ion, an analysis o he cu en s a e o he a ega ding he scope o his Mas e ’s Wo k will be made. The e o e, he concep s o an online co pus, compu e - media ed communica ion, na u al language p ocessing, and web sc aping will be explo ed. Addi ionally, ela ed wo k o his p ojec will be p esen ed. 2.1 online co pus Acco ding o he online Po uguese dic iona ies In opédia o he Po o Edi o a G oup and P ibe am, as well as he English online dic iona y Dic iona y, a co pus can be desc ibed as a compila ion o collec ion o documen s o in o ma ion abou a ce ain opic. I can come om a a ie y o sou ces, including books, newspape s, magazines, and online in o ma ion, and i can be used o linguis ic esea ch and asks in ol ing na u al language p ocessing. The online co pus included in he Ne Lang Co pus p ojec (Elias e al.,2021), which ocuses on online ha e speech, is an example o an online co pus. This p ojec aims o c ea e a co pus consis ing o ex s om a ious o ums (commen boa ds o news si es and social ne wo ks) ha ep esen he a iables ha will be analyzed. As such, hey in end o analyze and anno a e he linguis ic s uc u es and discu si e o ganiza ion o hose ex s. Finally, i aims o con ex ualize he da a ob ained in di e en eali ies wi hin he communi y o de ine awa eness and comba s a egies. 2.2 compu e -media ed communica ion Compu e -Media ed Communica ion (CMC) can be desc ibed as “ he p ocess by which people c ea e, exchange, and pe cei e in o ma ion using ne wo ked elecommunica ions sys ems ha acili a e encoding, ansmi ing, and decoding messages” (Romiszowski and Mason, 2013) and i can ake place ia echnology ools like compu e s, cellphones, and he in e ne . Email, ins an messaging, cha ooms, online o ums, social media, and ex messaging a e jus a ew o he many ac i i ies ha a e included in he scope o CMC. I is u ilized in dis an communica ion, g oup communica ion, and o he ypes o social in e ac ion, and 5 2.3. Web Sc aping 6 i enables communica ion be ween people who a e no physically p esen , which can be ad an ageous in ci cums ances whe e ace- o- ace in e ac ion is impossible o unpleasan . This Mas e ’s p ojec is ocused on his concep . Thus, he pu pose o his wo k is o c ea e an online co pus abou he con lic be ween Russia and Uk aine o ec ea e his social phenomenon. In o de o do so, i will be necessa y o ga he in o ma ion on he issue h ough social communica ion, and in he mode n e a, his communica ion occu s h ough bo h online newspape s and social ne wo ks, he la e o which is a mo e casual manne o dissemina ing in o ma ion. 2.3 web sc aping The echnique o Web Sc aping will be used o ex ac in o ma ion om online sou ces. Web sc aping is an old concep ha has exis ed nea ly since he in e ne was c ea ed. As a esul , his e m has been e e ed o by a ious names, including sc een sc aping and web ha es ing. This me hod en ails using so wa e o sc ip s o au oma ically ex ac da a om websi es. Tha is, i consis s o eques ing da a om a websi e’s se e , usually in he o m o HTML, pa sing ha da a, and hen ex ac ing he needed in o ma ion. To pa se he HTML code and ex ac he needed da a, a lib a y like Beau i ul Soup can be used (Mi chell,2018). This echnique allows he ex ac ion o a wide ange o da a ypes, including ex , images, ideos, and many o he s. I can be used o a ange o asks, including da a collec ion o esea ch and analysis, da a collec ion o machine lea ning, and da a collec ion o asks in ol ing na u al language p ocessing. Fo example, in he business in elligence a ea, i can be used o ex ac p oduc in o ma ion, such as p ices and e iews, om a ious online businesses o compa e p ices and ca y ou a ma ke analysis; in he a ea o da a science, i can be used o ex ac in o ma ion om social ne wo ks o do s a is ical and sen imen analysis and o he hings; e c (Khde ,2021). Gi en ha web sc aping is an old concep , he e a e se e al p ojec s ha ha e been de el- oped u ilizing his echnique. An example is he ollowing p ojec : “Ex ac ing and A chi ing Da a om Social Media o Suppo Cul u al He i age P ese a ion in Nine eh” (Rashid and Qasha,2022). This wo k in ends o collec in o ma ion abou Nine eh’s cul u al he i age h ough many social media si es. Wi h he p og amming language Py hon using he Sele- nium lib a y, he au ho s o his p ojec sc aped social media, such as Twi e , Ins ag am and YouTube, o ga he he needed in o ma ion. A e ha , hey cleaned he ex ac ed da a using Panda and egula exp ession lib a ies om Py hon. Finally, hey s o ed he p ocessed in o ma ion in a NoSQL da abase and amazon simple s o age s3. Ano he example is he p ojec “Applica ion o Web Sc aping and Google API Se ice o Op imize Con enience S o es’ Dis ibu ion” (Le and Pish a,2015), which is a wo k on he op imiza ion o con enience s o es in Japan. In his p ojec , he au ho s apply web sc aping 2.4. Na u al Language P ocessing 7 echniques o he NAVITIME websi e by using he p og amming language Ruby and i s ex e nal lib a y, Nokogi i. The p ojec “Co iFigh : A COVID-19 Resou ce T acke Websi e wi h Loca ion-wise Sea ch using Web Sc aping” is an example o a p ojec ha u ilizes web sc aping o collec in o ma- ion on a ecen global issue (Singh e al.,2022). This p ojec aims o c ea e a eal- ime COVID- 19 esou ce acke use - iendly websi e ha sc apes da a om social media pla o ms, such as Twi e , o display ele an in o ma ion abou he esou ces ela ed o COVID-19, gene al guidance in symp oms, ea men s, p ecau ions, e c. To do so, hey ex ac ed he da a om Twi e using he Twi e API and c ea ed he websi e whe e he on -end unc ionali ies a e implemen ed using HTML, CSS, and Ja aSc ip , and he back-end unc ionali ies ha e been coded in NodeJS. Finally, “Da a Analysis by Web Sc aping using Py hon” (Thomas and Ma hu ,2019) is ano he example o a p ojec ha uses web sc aping. This p ojec aims o ex ac in o ma ion om di e en sou ces, o example, e-comme ce si es such as Flipka and Amazon, using he web c awle Sc apy, which is a web-c awling amewo k de eloped in Py hon. Mo eo e , he in o ma ion is s o ed in a da abase and analyzed. 2.4 na u al language p ocessing Ano he concep add essed in his p ojec is Na u al Language P ocessing (NLP). NLP is a ield o a i icial in elligence and compu e science ha is ocused on p o iding compu e s wi h he capaci y o comp ehend w i en and spoken wo ds like humans (Kuma , 2013). Fo ins ance, NLP gi es compu e s he abili y o speak ecogni ion, ead ex , analyze i , de e mine he sen imen , and pinpoin he main pa s. Va ious examples o hese concep s include digi al assis an s, oice-ope a ed GPS sys ems, speech- o- ex dic a ion so wa e, da a and ex analysis, sea ch engines, and o he s. To assis he compu e in be e unde s anding he ex , NLP sys ems and models use a ious echniques om compu a ional linguis ics, compu e science, ma hema ics, and s a is ics. NLP is a concep ha has exis ed since he 1950s, and, as p e iously men ioned, he e a e mul iple echniques. As a esul , his concep and i s echniques a e he subjec s o se e al p ojec s, esea ch pape s, a icles, e c. Taking in o accoun he ela ed p ojec s ound simila o his Mas e ’s wo k and he aims o his p ojec , he ocus will be on he p ocess o ex p ep ocessing. The ex p ep ocessing p ocess consis s o cleaning and p epa ing he da a so ha i is in a o ma ha can be easily analyzed and p ocessed. I is an essen ial s ep in NLP. Typically, his p ocess in ol es a se ies o s eps ha can a y depending on bo h he ype o da a being p ocessed and he ask o which he da a will be applied. Fo example, he p ojec “P ep ocessing he In o mal Tex o e icien Sen imen Analysis” (Hemala ha e al.,2012) 2.5. Rela ed p ojec s 8 aims o p opose an e icien me hod o ex p ep ocessing p eceding he applica ion o any classi ica ion algo i hm o pe o m sen imen analysis. The ex unde wen he ollowing p ep ocessing s eps: emo al o URLs, elimina ion o ques ion wo ds like “how”, “wha ”, and o he s, and exclusion o special cha ac e s, e c. Ano he p ojec ha is an example o he usage o ex p ep ocessing is he ollowing: “Public Sen imen and Emo ion Analyses o Twi e Da a on he 2022 Russian In asion o Uk aine” (Ga cia and Cunanan-Yabu ,2022). This p ojec aims o s udy he public sen imen s o he in e na ional communi y abou he Russian in asion o Uk aine. They pe o med mul iple ex p ep ocessing echniques, such as emo ing special cha ac e s, URLs, noise le e s, and o he s, o p epa e he da a o emo ion and sen imen analysis. 2.5 ela ed p ojec s Because i has been a signi ican cause o ension and ins abili y in he egion o many yea s, he con lic be ween Uk aine and Russia is widely discussed on he in e ne . The e ha e been se e al inciden s as a esul o he wa , which has his o ical, poli ical, and economic oo s. Fu he mo e, he ongoing wa has had a p o ound impac on he global communi y, wi h mul iple na ions aking sides and imposing sanc ions on each o he . Conside ing he delica e na u e o he subjec , i has ga ne ed signi ican a en ion om he media. A e a sea ch, he ollowing p ojec abou he wa was ound: “Twi e Da ase on he Russo-Uk ainian Wa 1 ” (She so e al.,2022). This p ojec aimed o analyze he ends and opics discussed conce ning he con lic on he social media pla o m Twi e , analyzing he ends o use s, ha e speech, and p opaganda, among o he de ails. They used a da ase o 57.3million wee s om 7.7million people ha was c ea ed using he sea ch API o Twi e o ex ac old wee s, which enabled access o wee s om Feb ua y 24,2022, o example. Twi e ’s s eaming API was used o ga he mo e wee s based on ending hash ags associa ed wi h he speci ic opic o ga he e en mo e da a. To analyze he ends o use s, ha e speech, and p opaganda, among o he de ails, hey applied olume and sen imen analysis o he da ase c ea ed, mo e speci ically, a p e- ained mul i-language XLM-RoBERTa 2 model ha enables sen imen analysis o ze o-sho classi ica ion using use -de ined en i ies. To ga he mili a y in o ma ion, an XLM-RoBERTa model o ze o-sho classi ica ion using he label “mili a y” was used, and a cus om- ained NER (Named-en i y ecogni ion) model, using as a base spaCy’s NER model 3 and ained wi h he c ea ed da ase , was applied o iden i y en i ies like loca ion, pe son, weapon, e c., and so he wee s based on whe e hey we e loca ed in Uk aine. 1h ps://gi hub.com/alexd k14/RussoUk ainianWa _Da ase 2h ps://hugging ace.co/ icgalle/xlm- obe a-la ge-xnli-anli 3h ps://spacy.io/uni e se/p ojec / ideo-spacys-ne -model 2.5. Rela ed p ojec s 9 Ano he simila p ojec was ound, “A Reddi Da ase o he Russo-Uk ainian Con lic in 2022" (Zhu e al.,2022), which is an analysis o communi ies on Reddi abou he con lic whe e hey pe cei e he appa en p e e ence o he p o-Uk aine s ance in he “RU Con lic ” ca ego y, among o he aspec s. A e pe o ming a keywo d sea ch o b owse Reddi and na owing down hei selec ion o 12 sub eddi s, he au ho s o his p ojec buil a da ase comp ising use commen s abou he dispu e. The Pushshi .io API 4 was u ilized o ga he he da a. To pe o m he da a analysis and subsequen ly each a de ini i e conclusion, hey pe o med a ious cha ac e iza ion analyses on he daily olumes, pos s’ engagemen , ank, e m equency, linguis ic aspec s, and sen imen analysis o he ex . In he case o he sen imen analysis o he ex , hey eso ed o he Vade (Hu o and Gilbe ,2014) sen imen analysis ool. Ano he simila p ojec ha also ocuses on Reddi is “Happens ance: U ilizing Seman ic Sea ch o T ack Russian S a e Media Na a i es abou he Russo-Uk ainian Wa On Red- di ” (Hanley e al.,2022). This p ojec ocuses on he analysis o Russian p opaganda on he pla o m. To analyze he Russian p opaganda on new si es and social media, he au ho s o his p ojec u ilized sen ence-le el opic analysis using he MPNe 5 model on a icles. To ex ac opic keywo ds, he dimensionali y educ ion algo i hm UMAP (Bech e al.,2019), he densi y-based clus e ing algo i hm HDBSCAN (McInnes e al.,2017), and he class-based TF-IDF (Özgü e al.,2005) we e used. “Public Sen imen and Emo ion Analyses o Twi e Da a on he 2022 Russian In asion o Uk aine” (Ga cia and Cunanan-Yabu ,2022) is ano he simila p ojec ha was ound. This p ojec aims o in es iga e he sen imen s and emo ions o people a ound he wo ld and de e mine public opinion abou he 2022 con lic be ween Uk aine and Russia by analyzing da a ex ac ed om Twi e on he i s day o he con lic . Fo he collec ion, p ocessing, and analysis o da a, he au ho s employed he p og amming language R. To c ea e a co pus wi h he ex ac ed da a, hey used he Twi e API o ex ac he desi ed in o ma ion, collec ing a o al o 39525 wee s. Fo each wee , hey pe o med mul iple ex p ep ocessing echniques (such as emo ing special cha ac e s, URLs, noise le e s, e c) o p epa e he ex da a. To classi y he wee s in e ms o pola i y and emo ion, hey used a lexicon-based app oach using Bayesian classi ie s in he sen imen package. Ano he p ojec abou he con lic be ween Uk aine and Russia is i led “The Dynam- ics o Uk aine-Russian Con lic h ough he Lens o Demog aphically Di e se Twi e Da a” (Sazzed,2022). By examining commen s on he social media pla o m Twi e , his p ojec aims o comp ehend he opinions and eac ions o i s use s. To do his, hey used a da ase om he websi e Kaggle, a pla o m hos ing mul iple ypes o da ase s co e ing a ious opics. The da ase con ained abou 80000 wee s om use s in 7coun ies ac oss 4 con inen s. Con a y o he in o ma ion p o ided in he desc ip ion o he selec ed da ase , 4h ps://psaw. ead hedocs.io/en/la es / 5h ps://gi hub.com/mic oso /MPNe 2.5. Rela ed p ojec s 10 he au ho s ound se e al wee s ha had no hing o do wi h he wa be ween Russia and Uk aine, as well as wee s in a language o he han English, among o he hings. The e o e, o p epa e he da a o la e applica ion o NLP echniques such as sen imen analysis and opic modeling, he au ho s o his p ojec eso ed o ex p ep ocessing. La e , hey pe o med a ious ypes o s a is ical and seman ic analysis o in es iga e use sen imen s ac oss he e ie ed da a. They also explo ed he p esence o highly o ensi e wo ds and ph ases. Finally, o iden i y he mos discussed and p e alen opics, hey applied wo opic modeling app oaches: LDA and ans o me -based BERTopic. An addi ional simila p ojec was iden i ied: “Aspec based Sen imen and Emo ion Anal- ysis wi h ROBERTa, LSTM” (Si isha and Chandana,2022). This p ojec aims o unde s and people’s emo ions and opinions abou his wa by using mul iple pos s om people on Twi e . To accomplish i s goal, he au ho s collec ed 484221 wee s om a da ase called “Uk aine Con lic Twi e Da ase ” p esen on he websi e Kaggle. Then, o assess how Twi e discussions abou he Uk aine-Russia wa a ec public sen imen and so pe o m he sen imen analysis, hey applied a combina ion o he model Robe a wi h ABSA (Aspec Based Sen imen Analysis) and Long Sho -Te m Memo y (LSTM). Howe e , he objec i e o his Mas e ’s wo k is di e en . As p e iously s a ed, he pu pose is o econs uc he 2022 con lic be ween Uk aine and Russia. 3 PROPOSED APPROACH This chap e p esen s a p oposed app oach o he p ojec . The c ea ion o an online co pus ega ding he Uk ainian Wa is he aim o his p ojec . To accomplish his, we mus ex ac he equi ed da a, e ine i using na u al language p ocessing o il e ou he ele an in o ma ion, and hen s o e i in a eposi o y. A e comple ing all he p ocedu es, he online co pus c ea ion will begin. Finally, a websi e pla o m will be de eloped o use s o explo e he co pus. 3.1 sys em a chi ec u e As men ioned ea lie , he e a e a numbe o s eps ha need o be comple ed in o de o accomplish his p ojec , s a ing wi h he sea ch and selec ion o news pe inen o he heme, in o ma ion ex ac ion, e c., and concluding wi h he de elopmen o a web pla o m. The e o e, he main sys em’s a chi ec u e is depic ed in Figu e 1. A e he analysis o he ex s selec ed om online newspape s and social media and he selec ion o he iches , an ex ac o is applied ha ex ac s he desi ed in o ma ion ha is la e s o ed in a ile. In he ollowing s age, he in o ma ion is o ma ed and educed o only wha is essen ial using na u al language p ocessing echniques. Subsequen ly, he esul ing in o ma ion is con e ed o JSON o ma and hen s o ed in a eposi o y, which will hus o m he co pus. Finally, o explo e he knowledge con ained in he co pus, a que y p ocesso and Web pla o m will need o be de eloped. The que y p ocesso will ake eques s om use s o he online pla o m, sea ch he co pus, and hen deli e he eques ed da a. 11 3.2. Ex ac o s 12 Figu e 1: Sys em A chi ec u e 3.2 ex ac o s To cons uc he co pus, i was necessa y o selec and ex ac in o ma ion om online sou ces such as he Po uguese newspape s Público and Jo nal de Negócios, as well as he social media pla o m and online communi y Reddi . Público was selec ed as one o he newspape s o his wo k due o i s popula i y un il Feb ua y 2023, acco ding o a epo om he newspape i sel . Addi ionally, his newspape allows use s o sha e hei commen s on any a icles, which is aluable o his wo k since knowing people’s opinions on he subjec is impo an . This newspape p o ides se e al a icles; howe e , some a e exclusi e o subsc ibe s. Fo his wo k, only ee-access a icles we e conside ed. Meanwhile, Jo nal de Negócios was selec ed because i is a echnical newspape , in his case abou economics and inance, and he e o e has mo e a icles abou he economic impac o he con lic , o example. This newspape enables i s use s o engage in discussions h ough he commen s sec ion and p o ides a ange o a icles accessible ee o cha ge. Howe e , ce ain a icles a e ese ed o subsc ibe s only, o e ing exclusi e con en o his g oup. We ha e once again limi ed ou sea ch o only esou ces ha a e a ailable o ee. Finally, se e al online pla o ms we e analyzed, such as Facebook, Twi e , and Reddi , o collec he Po uguese’s opinion on he wa . As Facebook does no allow in o ma ion ex ac ion, his op ion was quickly disca ded. Be ween he wo esul ing op ions, Reddi was 3.2. Ex ac o s 13 chosen because, as no ed in he p ojec s ela ed o his wo k’s heme men ioned in Chap e 2, i is less used as a sou ce o in o ma ion. Conce ning news publica ions, i was necessa y o de elop an ex ac o capable o ex ac - ing he desi ed in o ma ion om a web page. Fo his pu pose, esea ch was conduc ed in he ield o Web Sc aping, he e m used o desc ibe he au oma ed p ocess o ga he ing s uc- u ed da a om he web o u he analysis. A simila wo k, he Ne lang p ojec 1 ” (Elias e al.,2021;Hen iques e al.,2019), which is no abou he con lic be ween Uk aine and Russia bu abou online speech, as p e iously men ioned, was used as a e e ence o he de elopmen o he ex ac o s. The e o e, a Py hon ex ac o was c ea ed ha sc apes he websi e o he Po uguese newspape Público, and ex ac s he a icle’s da a. To do ha , se e al Py hon lib a ies we e used. The ocus will be on he ollowing ones: Selenium 2 and Beau i ul Soup 3 . Selenium, which is an open-sou ce p og am, was used o au oma e he web b owse . A i s , we chose Ch ome as he p e e ed b owse o his p ojec . Howe e , due o incompa ibili ies wi h he compu e , we changed he chosen b owse o Fi e ox. The e o e, he selenium d i e used o open a b owse ins ance and un he sc ip c ea ed was he Fi e ox d i e , geckod i e . On he o he hand, Beau i ul Soup, which is a lib a y o pulling da a ou o HTML and XML iles, was used o ex ac he desi ed da a ela ed o he news publica ion, such as he a icle ex , au ho , i le, e c. To ex ac he commen s o a news publica ion, ano he ex ac o in Py hon, e y simila o he p e iously discussed one, was de eloped. The same Py hon lib a ies we e u ilized. In he case o he in o ma ion ha was aken om a news a icle om he online newspape , i is being s o ed in a JSON objec ha holds in o ma ion abou he a icle’s i le and sub i le, i s owne , he o al numbe o sha es, he da a o he a icle’s publica ion and i s ex ac ion, he language, he pla o m whe e i is a ailable, he a icle’s URL, he ex o he a icle, and he name o he newspape o which he a icle belongs. On he o he hand, de ails ega ding commen s ex ac ed om a newspape a icle a e sa ed in a lis o JSON objec s, whe e each objec has a numbe o key/ alue pai s ha s o e he de ails ega ding he commen . The ollowing in o ma ion is sa ed: commen ID, commen a o p o ile link, commen a o name, commen , and answe s o he commen a o ’s ema ks. I should be no iced ha he lis o JSON objec s in he ield co esponding o he commen s’ esponses ep esen s each commen ’s esponse. In Figu e 2and Figu e 3, an example o an a icle o be ex ac ed is displayed. I is possible o see he ex ha has o be ex ac ed as well as some o he commen s ha we e made when he a icle was published. Two examples o he JSON o ma in which he news and espec i e commen s a e ex ac ed a e also displayed. 1h ps://ne lang-co pus.ilch.uminho.p / 2h ps://pypi.o g/p ojec /selenium/ 3h ps://pypi.o g/p ojec /beau i ulsoup4/ 3.2. Ex ac o s 14 Figu e 2: Example o a news publica ion as an inpu o he ex ac o 4 GUCO, CORPUS CONSTRUCTION This chap e desc ibes he co pus cons uc ion p ocess. In his sec ion, he ex ac o s used o collec in o ma ion will be explained, as well as he o ma in which he in o ma ion was s o ed. Addi ionally, he ex -cleaning p ocess will be explained. To build he co pus, i was necessa y o selec and ex ac in o ma ion. A e esea ching he p ojec ’s heme, se e al ela ed p ojec s we e disco e ed, bu none o hem had he pu pose o building a co pus in he Po uguese language abou he con lic . The e o e, i was decided o c ea e a co pus wi h in o ma ion in he Po uguese language. To his end, as explained in Chap e 3, he Po uguese newspape s Público and Jo nal de Negócios we e chosen, as well as he social media pla o m and online communi y Reddi . 4.1 ex ac o s Conce ning he a icles om newspape s and hei commen s, i was necessa y o de elop an ex ac o capable o ex ac ing he desi ed in o ma ion om a web page. As p e iously men ioned, a simila wo k, he Ne lang p ojec 1 ”Elias e al. (2021); Hen iques e al. (2019), which is no abou he con lic be ween Uk aine and Russia bu abou online ha e speech, was used as a e e ence o he de elopmen o he ex ac o s. 4.1.1Público In his sec ion, he ex ac ion o a icles abou he con lic p esen in he Po uguese online newspape , “Público”, is explained. Fi s ly, an analysis o se e al news a icles was ca ied ou . I was obse ed ha nume ous a icles lacked commen s o had e y ew, while o he s had se e al, anging om dozens o hund eds o commen s. As a esul , i was decided o ex ac news ha had a leas en commen s o ex ac he ones ha had he mos impac on people. Only a icles wi hin he ime ame o Feb ua y 24,2022, o July 31,2023, we e conside ed. 1h ps://ne lang-co pus.ilch.uminho.p / 21 4.1. Ex ac o s 22 La e on, o ga he in o ma ion om “Público”, a Py hon ex ac o was c ea ed ha sc apes hei websi e and ex ac s he desi ed da a. The ex ac o c ea ed ecei es he URL o he chosen a icle as an a gumen and e u ns a JSON ile wi h he ex ac ed da a. To do ha , as men ioned in Chap e 3, se e al Py hon lib a ies we e used, such as Selenium and Beau i ul Soup. Selenium, which is an open-sou ce p og am, was used o au oma e he web b owse . On he o he hand, Beau i ul Soup, which is a lib a y o pulling da a ou o HTML and XML iles, was used o ex ac he desi ed da a ela ed o he news publica ion. Lis ing 4.1shows he ex ac o in Py hon used o collec in o ma ion ela ed o an a icle in he newspape “Público”. om selenium impo webd i e om selenium.webd i e . i e ox.se ice impo Se ice om selenium.webd i e . i e ox.op ions impo Op ions om selenium.webd i e impo Fi e ox om selenium.webd i e .common.by impo By om bs4 impo Beau i ulSoup impo ime impo json impo e impo sys op s = Op ions() bina y = "/us /bin/ i e ox" op s.bina y_loca ion = bina y op s.add_a gumen ("--headless") d i e = Fi e ox(se ice = Se ice("../d i e s/geckod i e "),op ions=op s) u l = sys.a g [1] d i e .ge (u l) ime.sleep(20) dados = d i e . ind_elemen ("id","s o y") h ml = dados.ge _a ibu e("inne HTML") soup = Beau i ulSoup(h ml, "h ml.pa se ") dic _news = {} i le = soup. ind("heade ", {"id":"s o y-heade "}) o i leN in i le. ind_all("h1", {"class":"headline s o y__headline"}): dic _news[' i le'] = i leN.ge _ ex ().s ip() subTi le = soup. ind("heade ", {"id":"s o y-heade "}) o subTi in subTi le. ind_all("di ", {"class":"s o y__blu b lead"}): dic _news['sub i le'] = subTi .ge _ ex ().s ip() Lis ing 4.1: Ex ac o o Newspape A icles coded in Py hon ( agmen ) 4.1. Ex ac o s 23 To ex ac he commen s o a news publica ion, ano he ex ac o in Py hon, e y simila o he p e iously discussed one, was de eloped. In his case, he ex ac o ecei es he ollowing da a as a gumen s: he link o he a icle, o e ie e he a icle da e ha is used in he gene a ed JSON ilename, and he link o he a icle commen s, ha is, he a icle link plus he ex “#commen s”. The same Py hon lib a ies we e used. The e ie ed da a is sa ed in a JSON ile. A his s ep o he p ocess, na u al language p ocessing echniques ha e no ye been applied; he e o e, he ex ac ed da a is un o ma ed. In he case o in o ma ion ex ac ed om an a icle, i is being s o ed in a JSON objec ha holds in o ma ion abou he a icle’s i le ( i le) and sub i le (sub i le), i s owne (owne ), he o al numbe o sha es (sha es), he da a o he a icle’s publica ion (da ePos ed) and i s ex ac ion (da eEx ac ion), he language in which i is w i en (language), he pla o m whe e i is a ailable (pla a o m), he a icle’s URL (u l), he ex o he a icle (pos Tex ), he name o he newspape o which he a icle belongs (nameNewspape ), and he numbe o commen s (commen s). Lis ing 4.2shows an example o he in o ma ion ex ac ed in JSON o ma om an a icle om he newspape “Público”: 1{ 2" i le":"Rússia assume p esidência do Conselho de Segu ança,a 'pio anedo a' de 1 de Ab il", 3"sub i le":"Lide ança o a i a do ó gão das Nações Unidas pe ence ago a a Mosco o. Uc ânia c i ica e EUA dizem que não há o ma legal de impedi a Rússia de assumi o ca go.", 4"owne ":[ 5"Paulo Na igão Reis", 6"Agências" 7], 8"sha es":"21 sha es", 9"da ePos ed":"1 de Ab il de 2023,13:56", 10 "da eEx ac ion":"2023-04-11", 11 "language":"p ", 12 "pla a o m":"Newspape ", 13 "u l":"h ps://www.publico.p /2023/04/01/mundo/no icia/ ussia-assume- p esidencia-conselho-segu anca-pio -anedo a-1-ab il-2044664", 14 "pos Tex ":[ 15 "A Rússia assumiu es e sábado a p esidência o a i a do Conselho de Segu ança das Nações Unidas, o nando-se no p imei o país a di igi o ó gão cujo P esiden e,Vladimi Pu in,é al o de um mandado in e nacional de cap u a po alegados c imes de gue a.","Leia os 4.1. Ex ac o s 24 a igos que quise ,a é ao im","Com uma assina u a mensal em acesso ilimi ado a odos os con eúdos e cancela quando quise ","`` É a pio anedo a do dia das men i as do mundo'',disse o minis o dos Negócios Es angei os da Uc ânia,Dmi o Kuleba,comen ando o ac o de Mosco o assumi a p esidência do ó gão da ONU no dia 1 de Ab il.", 16 ... 17 ], 18 "nameNewspape ":"Público", 19 "commen s":"10 commen s" 20 } Lis ing 4.2: Example o an ex ac ed a icle om “Público” in JSON o ma On he o he hand, de ails ega ding commen s ex ac ed om a newspape a icle a e sa ed in a lis o JSON objec s, whe e each objec has a numbe o key/ alue pai s ha s o e he de ails ega ding a commen . The ollowing in o ma ion is sa ed: he id o he commen (ID), he commen a o p o ile link (P o ile Link), he commen a o anking (Ranking), he commen a o ’s name (Name), he ex o he commen (Commen ), and he answe s o he commen a o ’s ema ks (Responses). I should be no iced ha he lis o JSON objec s in he ield co esponding o he commen s’ esponses ep esen s each commen ’s esponse. Lis ing 4.3shows an example in JSON o ma o some commen s ex ac ed om a newspape a icle pos ed in “Público”: 1{ 2"ID":"3b55697d- 38c-4 2 -d8d1-08db04c372ee", 3"P o ile Link":"h ps://www.publico.p /u ilizado /pe il/05d8 635-cc c-41 ca-812b-9 3e6b9 c 77", 4"Ranking":"In luen e ", 5"Name":"Ca a ina Pomba ", 6"Commen ":" n Se á que o Sec e á io-Ge al das Nações Unidas não pode ia exp imi -se - e u gen emen e,pois já es amos no dia 2 de ab il - um an o ou quan o o a 7da caixa? ... n ", 8"Responses":[ 9{ 10 "ID":"9d 362a4-3bcc-4a2d-2bd5-08db04d78ace", 11 "P o ile Link":"h ps://www.publico.p /u ilizado /pe il/1 85203e -62da-414d-a56a-e93189dcc5be", 12 "Ranking":"Inician e ", 4.1. Ex ac o s 25 13 "Name":"joao eis ", 14 "Commen ":" n Ca a ina a subse iência do Sec e á io-Ge al es á em consonância com a hipoc isia gene alizada dos países que compõem a o ganização ao eme e em- se ao silêncio nos discu sos e nas posições ace ca da in asão da Uc ânia. nSe a hones idade in elec ual osse uma ca ac e í s ica desses países jamais um país in aso e ia o di ei o a pa icipa na o ganização,quan o mais lide á-la... n " 15 }, 16 { 17 "ID":"d1a04d53-106d-4160-4b 6-08db256e9d62", 18 "P o ile Link":"h ps://www.publico.p /u ilizado /pe il/05d8 635- cc c-41ca-812b-9 3e6b9 c 77", 19 "Ranking":"In luen e ", 20 "Name":"Ca a ina Pomba ", 21 "Commen ":" n Conco do 22 consigo,João Reis. n " 23 } 24 ] 25 } Lis ing 4.3: Example o some commen s ex ac ed om a newspape a icle pos ed in “Público” is p esen ed in JSON o ma A o al o 378 a icles and 11334 commen s we e ex ac ed. 4.1.2Jo nal de Negócios In his sec ion, he ex ac ion o a icles abou he con lic p esen in he Po uguese online newspape “Jo nal de Negócios” is explained. In his case, he news selec ion p ocess used was qui e simila o ha used in he newspape “Público”. Howe e , i was ound ha in e es ing news, such as he impac o he con lic on he coun ies’ economies, had ew commen s. The e o e, we decided o educe he minimum numbe o commen s o i e and accep excep ions, ha is, news wi h less han i e commen s bu wi h in o ma ion ha was pe sonally conside ed ele an . As in he case o he newspape “Público”, only a icles wi hin he ime ame o Feb ua y 24,2022, o July 31,2023, we e conside ed. 4.1. Ex ac o s 26 To ga he da a om he selec ed a icles, an ex ac o mi o ing he one used o “Público” a icles was de eloped. Conce ning he ex ac ion o commen s, ano he ex ac o like he one designed o “Público” commen s was c ea ed. In bo h cases, because he page s uc u e o he “Jo nal de Negócios” newspape is di e en , some changes o he ex ac o s we e made. A his poin , he ex ac ed da a is sa ed in JSON o ma and NLP echniques ha e no ye been applied, he e o e ex cleaning has no ye been done. Because o ha , he in o ma ion is no co ec ly o ma ed. Lis ing 4.4illus a es he JSON- o ma ed in o ma ion ex ac ed om a newspape a icle ea u ed in “Jo nal de Negócios”: 1{ 2" i le":"Economia mundial a caminho da es agnação nos p óximos cinco anos, a isa FMI", 3"sub i le":"Ins i uição apon a pa a um c escimen o in e io a 3% pa a es e ano. A con i ma em-se as p e isões de médio p azo,ou seja,num pe í odo de cinco anos,se ão as pio es das úl imas ês décadas.", 4"owne ":"Paulo Ribei o Pin o", 5"da ePos ed":"06 de Ab il de 2023,15:21", 6"sha es":"n/a", 7"da eEx ac ion":"2023-04-21", 8"language":"p ", 9"pla a o m":"Newspape ", 10 "u l":"h ps://www.jo naldenegocios.p /economia/conjun u a/de alhe/ economia-mundial-a-caminho-da-es agnacao-nos-p oximos-cinco-anos-a isa- mi", 11 "pos Tex ":[ 12 "Apesa da g ande obus ez do me cado de abalho e da esis ência do consumo p i ado,a economia mundial de e á c esce es e ano abaixo dos 3% e,no médio p azo, e um compo amen o aco,só is o na d écada de 1990.","Em esumo é es a a ideia cen al deixada es a quin a- ei a pela di e o a-ge al do FMI,K is alina Geo gie a,no discu so 13 que ma ca a abe u a dos encon os do Banco Mundial e do Fundo Mone á io,em Washing on,nos Es ados Unidos."," "Depois de uma o e ecupe ação em 2021 chegou um choque se e o 14 na gue a na Uc ânia de amplas consequências - o c escimen o global em 2022 desceu pa a p a icamen e me ade,de 6,1% pa a 3,4% ",começou po lemb a a che e do FMI. Mas o ab andamen o económico man e e-se 4.1. Ex ac o s 27 es e ano, isou. E apesa de "um me cado de abalho esilien e al como o consumo nas economias a ançadas ",da ecupe ação da China de ido à eabe u a,a economia mundial de e á "c esce menos de 3% em 2023 ",apon ou. Uma p e isão em linha com as p ojeç ões de janei o de 2,9%.", 15 ... 16 ], 17 "nameNewspape ":"Jo nal de Negócios", 18 "commen s":" n 10 n commen s" 19 } Lis ing 4.4: Example o an ex ac ed a icle om “Jo nal de Negócios” in JSON o ma Howe e , in his case, he in o ma ion ex ac ed om he commen s is sligh ly di e en since he commen s in “Jo nal de Negócios” ha e a di e en s uc u e. Tha is, he mos - o ed commen appea s highligh ed a he beginning. Howe e , in ha case, i s numbe o likes and dislikes, as well as i s esponses, a e no p o ided he e. Then, he es o he commen s a e p esen ed, and he mos - o ed commen appea s again in i s co ec o de wi h all he in o ma ion a ailable. In his case, he ollowing in o ma ion is being sa ed: he id o he commen (ID), he i le o he commen ( ha is, whe he o no i is he mos o ed commen ) (Commen ’s Ti le), he commen a o ’s name (Name), he commen ’s da e (Commen Da e), he numbe o likes and dislikes o he commen (Numbe o Likes and Numbe o dislikes), he ex o he commen (Commen ), and he answe s o he commen a o ’s ema ks (Responses). I should be no iced ha he lis o JSON objec s in he ield co esponding o he commen s’ esponses ep esen s each commen ’s esponse. Lis ing 4.5shows commen s om a newspape a icle om “Jo nal de Negócios”, p e- sen ed in JSON o ma : 1[ 2{ 3"ID":"n/a", 4"Commen 's Ti le":"Comen á io mais o ado", 5"Name":"Anónimo", 6"Commen Da e":"Há 55 dias e 7 ho as", 7"Commen ":" "Ba il de pe óleo pode subi aos 100 dóla es ":Com o co e anunciado pela OPEP+ na p odução de pe óleo à escala global, na o dem de de 1,16 milhões de ba is po dia,a subida de p eço daquele combus í el é uma endência al amen e p o á el. Alguns analis as ale am mesmo pa a a e en ualidade de os cus os subi em 4.1. Ex ac o s 28 de al o ma que alcancem os 100 dóla es po ba il. " - h ps:// jo naleconomico.p /no icias/ba il-de-pe oleo-pode-subi -aos-100- dola es-es es-sao-os-paises-que-se iam-mais-a e ados-1015600", 8"Responses":[] 9}, 10 ... , 11 { 12 "ID":"2777102", 13 "Name":"Anónimo", 14 "Commen Da e":"Há 55 dias e 7 ho as", 15 "Numbe o Likes":"3", 16 "Numbe o dislikes":"0", 17 "Commen ":" "Ba il de pe óleo pode subi aos 100 dóla es ":Com o co e anunciado pela OPEP+ na p odução de pe óleo à escala global, na o dem de de 1,16 milhões de ba is po dia,a subida de p eço daquele combus í el é uma endência al amen e p o á el. Alguns analis as ale am mesmo pa a a e en ualidade de os cus os subi em de al o ma que alcancem os 100 dóla es po ba il. " - h ps:// jo naleconomico.p /no icias/ba il-de-pe oleo-pode-subi -aos-100- dola es-es es-sao-os-paises-que-se iam-mais-a e ados-1015600", 18 "Responses":[ 19 { 20 "ID":"2777127", 21 "Name":"Anónimo", 22 "Commen Da e":"Há 55 dias e 4 ho as", 23 "Numbe o Likes":"1", 24 "Numbe o dislikes":"0", 25 "Commen ":"Pensem é em acaba ou eduzi ao máximo possí el es e canc o! ", 26 "Responses":[] 27 } 28 ] 29 }, 30 ... , 31 { 32 "ID":"2777095", 33 "Name":"Joaquim Schiede da Sil a", 34 "Commen Da e":"Há 15 dias e 4 ho as", 4.1. Ex ac o s 29 35 "Numbe o Likes":"3", 36 "Numbe o dislikes":"2", 37 "Commen ":"En ao iniciam uma gue a com a Russia es a am espe a de que?", 38 "Responses":[ 39 { 40 "ID":"2777128", 41 "Name":"Anónimo", 42 "Commen Da e":"Há 15 dias e 1 ho as", 43 "Numbe o Likes":"0", 44 "Numbe o dislikes":"0", 45 "Commen ":"A in asão,os c imes e as a ocidades não 46 o e ecem dú idas!", 47 "Responses":[] 48 } 49 ] 50 } 51 ] Lis ing 4.5: Example o commen s ex ac ed om a newspape a icle om “Jo nal de Negócios” in JSON o ma Ul ima ely, a o al o 730 a icles and 6119 commen s we e collec ed. 4.1.3Reddi This sec ion explains he p ocess o ex ac ing pos s and hei commen s om he pla o m “Reddi ”. To ex ac he in o ma ion, wo ex ac o s we e de eloped in he p og amming language Py hon. Fo each ex ac o , he Py hon lib a y PRAW 2 (Py hon Reddi API W appe ) was used, as i p o ides an in e ace o accessing he Reddi API. PRAW is buil on op o he Reddi API and allows use s o connec hei Py hon code o Reddi . To ha e access o he da a, i was necessa y o c ea e a Reddi accoun and hen go o eddi .com/p e s/apps and c ea e a Reddi applica ion. Following ha , a sc ape ins ance was c ea ed using he Reddi class om he PRAW lib a y. Reddi is a pla o m ull o in o ma ion om a ious use s om di e en pa s o he wo ld. I is di ided in o smalle communi ies, called “sub eddi ”. The e is a communi y o any hing and e e y hing. Fo his wo k, we decided o ex ac in o ma ion in Po uguese 2h ps://p aw. ead hedocs.io/en/s able/ 4.1. Ex ac o s 30 abou he con lic be ween he wo coun ies; he e o e, only in o ma ion p esen in he communi y “po ugal” was conside ed. As such, in each ex ac o , he sub eddi om which he in o ma ion would be ex ac ed (“ /po ugal”) was de ined. To only sc ape pos s ela ed o he con lic be ween Uk aine and Russia, he sub eddi was sea ched using he keywo ds “Uk aine” and “Russia”. The esul s we e limi ed o he pe iod be ween Feb ua y 24,2022, and July 31,2023. The ex ac ed da a is cu en ly being sa ed in JSON o ma . A his poin , since NLP echniques ha e no ye been applied, he ex cleaning has no ye been done. Because o ha , he ex ac ed in o ma ion is un o ma ed. In he case o in o ma ion ex ac ed om a pos , i is being s o ed in a JSON objec ha holds in o ma ion abou he pos ’s i le ( i le), i s sco e (sco e), i s id (id), he pos ’s URL (u l), he numbe o commen s (comms_num), i s imes amp (c ea ed), and i s desc ip ion (body). Lis ing 4.6p esen s an example o a pos ex ac ed om he pla o m “Reddi ” in JSON o ma : 1{ 2" i le":"Po ugal p epa a en io de qua o anques pa a a Uc ânia", 3"sco e":"108", 4"id":"10kw u0", 5"u l":"h ps://www. eddi .com// /po ugal/commen s/10kw u0/ po ugal_p epa a_en io_de_qua o_ anques_pa a_a/", 6"comms_num":"104", 7"c ea ed":"2023-01-25 11:44:41", 8"ex ac ed":1693534312, 9"body":"Po ugal p epa a en io de qua o anques Leopa d 2 pa a a Uc ânia, depois de a Alemanha e cedido inalmen e à p essão dos aliados a dado luz e de ao en io daqueles ca os de comba e. Re i a-se que Po ugal em 37 daqueles blindados,12 dos quais o almen e ope acionais. n nAcei am-se apos as:os 4 pa a en ia azem pa e dos 12 ope acionais,ou ai se como com helicóp e os? n nFon e:[h ps:// www.cmjo nal.p /mundo/de alhe/po ugal-p epa a-en io-de-qua o- anques- pa a-a-uc ania](h ps://www.cmjo nal.p /mundo/de alhe/po ugal-p epa a- en io-de-qua o- anques-pa a-a-uc ania)" 10 } Lis ing 4.6: Example o a pos ex ac ed om he pla o m “Reddi ” in JSON o ma On he o he hand, de ails ega ding commen s ex ac ed om a pos a e sa ed in a lis o JSON objec s, whe e each objec has many key/ alue pai s ha s o e he de ails ega ding he commen . The ollowing in o ma ion is sa ed: he commen id (commen _id), he id o he 4.3. Da a P ocessing 37 numbe o commen s lea ing i wi h jus he nume ic alue. Bo h he a icle ex ac ion da e and publica ion da e we e changed o ollow he gene al JSON o ma p e iously c ea ed (“yyyy-mm-dd hh:mm”). In he case o he in o ma ion ex ac ed ega ding he commen s o each a icle, an analysis was ca ied ou and i was e i ied, once again, he exagge a ed exis ence o line change cha ac e s in a ow; he exis ence o commen s ha can be classi ied as SPAM, such as “O Google pagou po abalho online diá io de casa de 400 a800 dóla es”,“Boas no ícias pa a odos os meus amigos ..... Googleee pagando”, and “O Google pagou po cada semana de abalho on-line em casa de 6.000 a10.000 eu os”; and he p esence o he cha ac e “ ” behind he cha ac e co esponding o he quo a ion ma ks. As such, a Py hon sc ip was de eloped ha made he men ioned changes. In addi ion, i also elimina ed epea ed commen s and adap ed he o ma o he commen ’s da e o he gene al o ma c ea ed. Finally, his sc ip adap ed he o ma o he ex ac ed in o ma ion o he gene al o ma . Using a me hodology simila o ha o he “Público” newspape , a inal sc ip was gene a ed o combine he in o ma ion om he a icles and hei co esponding commen s in o a single JSON ile. 4.3.3Reddi In his sec ion, an explana ion o he ex cleaning p ocess applied o all he in o ma ion ex ac ed om he pla o m “Reddi ” will be p o ided. As an ini ial s ep, an analysis o he ex ac ed in o ma ion was conduc ed. Du ing his analysis, pos s un ela ed o he con lic we e iden i ied and, consequen ly, emo ed. Subsequen ly, wo dis inc sc ip s we e c ea ed: one o he in o ma ion linked o pos s e ie ed using he keywo d “Uc ânia”, and he o he o pos s ex ac ed wi h he keywo d “Russia”. These sc ip s il e ed ou pos s lacking speci ic keywo ds in hei i les o body, such as “gue a”, “con lic o”, “in asao”, “con li o”, “ba alha”, “con on o”, “ ladimi ”, “pu in”, “zelensky”, “ olodymy ”, “comba e”, “dispu a”, “ e ugiado”, “adesao a ue”, “adesao a uniao eu opeia”, “p o-uc ania”, and “p o- ussia”. Following a ho ough e iew o he ou comes, non-con lic - ela ed pos s and hose wi h i e o ewe commen s we e elimina ed. Addi ionally, since he sea ch using he keywo ds “Russia” and “Uc ânia” was conduc ed wi hin he same sub eddi , a Py hon sc ip was de eloped o ensu e he absence o duplica e pos s. In he case o edundan pos s, one o he duplica es was emo ed. While analyzing he pos s, i was also obse ed ha he ex ac ion o he language in which he pos s and commen s we e w i en, along wi h he iden i ica ion o he pos ’s au ho , was no done. Fo his eason, wo sc ip s in Py hon we e c ea ed, one o ex ac he language used in he pos s and commen s, and he o he o ex ac he au ho s o he pos s. 4.3. Da a P ocessing 38 In he case o he language iden i ica ion, he nl k lib a y was used o do ex classi ica ion and guess he language o he ex . A e ha , a sc ip in py hon o clean he pos s and adap hei o ma o he gene al o ma was c ea ed. This sc ip emo es he slashes be o e he quo a ion ma ks; change he quo a ion ma ks o “ ’ ’ ”; modi y he da e o he pos and when i was ex ac ed o he co ec o ma (“yyyy-mm-dd hh:mm”) ; emo e he ex a slash in he pos ’s URL, emo e special cha ac e s and con e he body o he pos o a lis o s ings. A e wa ds, he commen s on he pos s we e analyzed. Fi s ly, i was ound ha he commen s we e no ex ac ed in he co ec o de . Also, in se e al commen s, a slash appea ed be o e he quo a ion ma ks, he e we e se e al line change cha ac e s in a ow, he e was he p esence o special cha ac e s, and an ex a slash in he commen URL. To his end, a Py hon sc ip was c ea ed ha makes he men ioned changes, adap s he commen o ma o he gene al o ma , and o de s he commen s. Finally, a sc ip in Py hon was de eloped o me ge he in o ma ion om he pos s and hei commen s in o a single JSON ile. 5 GUCO, PLATFORM A e ex ac ing and applying na u al language p ocessing echniques, he nex s eps we e o s o e he in o ma ion and p o ide use s wi h a pla o m o explo e he knowledge con ained in he co pus. Hence, his chap e desc ibes he de elopmen o he p ojec websi e c ea ed called “GUCO”. In his Chap e , he backend and on end o he websi e will be explained. Addi ionally, he mockups designed o he websi e will be p esen ed, as well as i s ul ima e design. Mo eo e , we will o e guidance on e ec i ely na iga ing and u ilizing he pla o m. 5.1 mockups In his sec ion, he mockups designed o he GUCO si e will be p esen ed. Be o e building he websi e, mockups we e c ea ed o o malize he idea o how he in o ma ion should be p esen ed o he use s. To his end, he ool Figma1was used. Conside ing he heme o his p ojec , i was decided ha i would be bes o design a websi e ha is isually simple and use - iendly. As such, we decided o ollow he same idea on almos e e y page: a ixed side menu on he le side o he page, and he igh side o he page changes depending on he menu op ion ha was selec ed. Figu e 4 ep esen s he home page o he web pla o m. On he home page, he idea was o display he abs ac and in oduc ion o he p ojec on he igh side o he page and ha e a ixed menu on he le side o he page in which he use can choose whe e o go. 1h ps://www. igma.com/ 39 5.1. Mockups 40 Figu e 4: (Mockup) GUCO Websi e: Home Page Figu e 5displays he “P ojec ” op ion on he menu. The menu p o ides he use s wi h mul iple op ions, including he p ojec ’s in oduc ion, objec i es, sys em a chi ec u e, and de ails abou he ex ac o s c ea ed o he p ojec . Figu e 5: (Mockup) GUCO Websi e Menu: P ojec Page Op ion 5.1. Mockups 41 Figu e 6shows he “Ex ac o s” op ion on he menu. This p o ides mul iple op ions o he use s, including in o ma ion abou he ex ac o s c ea ed o ex ac ing da a om bo h newspape s, “Público” and “Jo nal de Negócios”, as well as he ex ac o c ea ed o he pla o m “Reddi ”. Figu e 6: (Mockup) GUCO Websi e Menu: Ex ac o s Page Op ion Figu e 7po ays he “Co pus” page. On his page, he use s can access he co pus. To do ha , hey need o ill ou he p o ided o m ha allows hem o selec he pla o m hey wan o analyze he in o ma ion om (Público, Jo nal de Negócios, o Reddi ) as well as he da e ange o ha in o ma ion. To make oom o hese op ions, he side menu was eloca ed o he op o he page. Figu e 7: (Mockup) GUCO Websi e Co pus: Co pus Page 5.1. Mockups 42 Figu e 8 ep esen s he page o he esul o a sea ch in he co pus. On his page, all he i les o he a icles o pos s esul ing om he sea ch in he co pus a e p esen ed o he use s. Figu e 8: (Mockup) GUCO Websi e Co pus: Lis o esul s A e selec ing one o he esul s om he sea ch on he co pus, all he in o ma ion abou an a icle o pos becomes a ailable o he use s. Figu e 9illus a es he p esen a ion o ha in o ma ion. 5.2. Backend 43 Figu e 9: (Mockup) GUCO Websi e Co pus: Resul Page Mockups we e no c ea ed o he emaining pages as hey would be iden ical o hose c ea ed o he home page, o example. 5.2 backend In his sec ion, he da abase used and how he da a is being s o ed will be discussed, as well as he websi e backend de elopmen p ocess. 5.2. Backend 44 Fo his p ojec , i was decided o use a non- ela ional da abase due o i s high lexibil- i y, high pe o mance, and low complexi y. As such, a MongoDB 2 da abase was chosen. MongoDB is a documen da abase in which da a is s o ed in collec ions as BSON docu- men s, which a e o ma ed simila ly o JSON. The e o e, he co pus in o ma ion is s o ed in a da abase called “GUCOdb” and is di ided in o h ee di e en collec ions. Tha is, all a icles and hei commen s ex ac ed om he “Público” newspape a e s o ed in a collec ion called “publico”. On he con a y, all a icles and hei commen s ex ac ed om “Jo nal de Negócios” a e sa ed in a collec ion called “jo nal_negocios”. Finally, all pos s and espec i e commen s ex ac ed om he “Redd i ” pla o m a e s o ed in a collec ion called “ eddi _pos s”. Th oughou he p ocess o s o ing he ex ac ed in o ma ion, we encoun e ed a challenge. When a emp ing o impo a JSON ile con aining da a ela ed o a Reddi pos , an e o occu ed du ing he da abase impo . This issue appea ed due o he p esence o nume ous nes ed commen s, exceeding MongoDB’s limi o 100 nes ed BSON documen s. To add ess his limi a ion, we op ed o G idFS, a speci ica ion o s o ing and e ie ing iles ha su passes MongoDB’s cons ain s. G idFS di ides he ile in o pa s, o chunks, and s o es each chunk as a sepa a e documen . To s o e each documen , wo collec ions we e used. One s o es he ile chunks, and he o he s o es he documen me ada a. A e s o ing he in o ma ion, he p ocess o c ea ing a Web pla o m in which use s would be able o explo e he co pus was ini ia ed. To do so, he Py hon amewo k Flask 3 was used. Flask is a amewo k designed o he Py hon p og amming language, used o build web applica ions. In his p ojec , he Flask applica ion (app) is di ided in o many sc ip s and olde s. The “app.py” sc ip is used o c ea e he Flask app. Also, in his sc ip , he bluep in c ea ed in “ iews.py” is impo ed and egis e ed. In he “ iews.py” sc ip , a bluep in is c ea ed o o ganize he ela ed iews, which a e he unc ions used o espond o he eques s o he applica ion. Addi ionally, in his sc ip , he connec ion o he da abase is made. In he olde called “ empla es”, all he HTML iles used o ende he in o ma ion on he b owse a e loca ed. Finally, in he olde called “s a ic”, he CSS ile used o de ine he isual p esen a ion o he HTML elemen s is loca ed. The ocus will be on he “ iews.py” sc ip and he unc ions c ea ed esponsible o handling HTTP eques s and ende ing esponses. Fi s ly, a unc ion o de ine he ou e o he home page was c ea ed. This ou e handles he eques s o he URL “/”. I dynamically e ie es a lis o images om he “images” olde wi hin he “s a ic” di ec o y, c ea es a lis o cap ions and c edi s o each image and subsequen ly ende s he “base.h ml” empla e by passing he images, cap ions, and c edi s as a iables. Then, a unc ion was c a ed o es ablish he ou e o he eam page. This pa icula ou e manages eques s o 2h ps://www.mongodb.com 3h ps:// lask.palle sp ojec s.com/en/3.0.x/ 5.3. F on end 45 he URL “/ eam” and ende s he “ eam.h ml” empla e. Subsequen ly, ano he unc ion was c ea ed o de ine he ou e o he publica ions page. I esponds o eques s om he URL “/publica ions” and ende s he “publica ions.h ml” empla e. A e ha , a new ou e o he in oduc ion page o he p ojec was implemen ed. I handles he eques om he URL “/in o” and ende s he “in o.h ml” empla e. Following ha , a ou e o he objec i es o he p ojec page was c ea ed. I handles he eques om he URL “/aims” and ende s he “aims.h ml” empla e. Fo he pages p esen ing he ex ac o s designed o collec in o ma ion om h ee dis inc sou ces, many ou es we e es ablished. Each ou e handled he eques s om di e en URLs, such as “/ex _pub”, “/ex _jn”,“/ex _ ed”, and ende ed he “ex _pub.h ml”, he “ex _jn.h ml”, and he “ex _ ed.h ml” empla es espec i ely. Addi ionally, he ou e o he co pus page was c ea ed. I handles a “GET” o “POST” eques o he URL “/co pus”. Fi s , i es s i he eques is yped “POST”. I ue, i s o es he in o ma ion ha was inse ed om he use in he o m and uses i o sea ch he da abase o he equi ed a icles o pos s. Then, i ende s he “co pus.h ml” empla e by passing mul iple a iables such as he inal esul o he sea ch on he co pus, he lis o chosen sou ces, he lis o e ms inse ed, and bo h he s a and end da es selec ed by he use . Finally, a ou e o he esul page o a sea ch on he co pus was c ea ed. I handles a eques o he URL “/elem/<sou ce>/<elem_ i le>”. Gi en he sou ce (“Publico”, “Jo nal de Negócios” o “Reddi ”) and he i le o he esul , he unc ion sea ches on he da abase and ende s he empla e “elem2.h ml” wi h all he in o ma ion ela ed o he esul a icle o pos . 5.3 on end This sec ion aims o p o ide an explo a ion o he on end de elopmen p ocess o he en i e websi e. I will de ail he echnologies employed in he p ocess and explain he websi e’s design. On his p ojec ’s websi e, he on end was buil wi h HTML (Hype ex Ma kup Lan- guage), CSS (Cascading S yle Shee s), and Ja aSc ip . HTML is a language ha speci ies web page s uc u e. CSS is a language used o desc ibe he web page’s appea ance. Ja aSc ip is used in web de elopmen o speci y he beha io o he web page. To acili a e he c ea ion o esponsi e and aes he ically pleasing pages, Boo s ap 4 , which is an open-sou ce amewo k, was used. To be mo e p ecise, he SB Admin 5 empla e was used as he basis o he websi e. As men ioned be o e, he Flask websi e is di ided in o olde s and sc ip s. Inside he “ empla es” olde , he HTML sc ip s used o ende he in o ma ion on he b owse a e 4h ps://ge boo s ap.com/ 5h ps://s a boo s ap.com/ empla e/sb-admin 5.3. F on end 46 loca ed. Howe e , inside he “s a ic” olde , he CSS ile called “s yles.css”, is used o de ine he isual p esen a ion o he HTML elemen s. As p esen ed in he mockups, he design o mos pages is qui e simila . Gene ally, each page consis s o a ixed side menu, p esen on he le side o he page, which has he name o he p ojec and he websi e c edi s, as well as bu ons o he possible pages o he websi e: Home, P ojec , Co pus, Publica ions, and Team. The P ojec bu on is di ided in o se e al op ions: he p ojec ’s in oduc ion, i s objec i es, he sys em a chi ec u e, and in o ma ion abou he many ex ac o s c ea ed. Usually, on he igh side o he page he i le o he p ojec is displayed along wi h o he in o ma ion depending on he page he use is in. Howe e , du ing he de elopmen o he on end, i was no ed ha some changes o he design we e needed. A i s , on he co pus page, he le -side menu was designed o be on he op o he page and he esul s o a sea ch in he co pus would be displayed on he igh side o he co pus sea ch o m. None heless, i was concluded ha he in o ma ion shouldn’ be p esen ed ha way because i would no be isually appealing as he e could be a lo o in o ma ion in he same space gi en he numbe o esul s o a sea ch on he co pus. Mo eo e , in he mockups, upon he use ’s click on each esul , all pe inen in o ma ion would be showcased on he same page. I became clea ha displaying mul iple esul s and hei in o ma ion in he same page would no be a good idea as i would cause he page o g ow signi ican ly in size. The e o e, i was decided ha on he co pus page, he le -side menu would s ay in i s place like on he o he pages, and only he i les o he esul s o a sea ch on he co pus would be displayed on ha same page. Howe e , he in o ma ion abou each esul would be sha ed on a new web page, a e clicking on i s i le’s bu on. Finally, he colo scheme o he websi e was also changed o a da ke colo as he ligh colo was no as appealing and was oo in ense, which made he si e’s isibili y di icul on some de ices. Addi ionally, ano he ield was added o he o m on he co pus page ( he “Selec e ms”) so he use could il e he in o ma ion o he co pus and only ha e access o he a icles o pos s wi h he e ms inse ed p esen on hei i le. Due o he iden ical design ac oss mos pages, an HTML sc ip wi h he common in- o ma ion was c ea ed o educe code leng h and a oid edundancy. In his sc ip , called “base.h ml”, in o ma ion ela ed o he ixed le -sided menu is ende ed, and he i le o he p ojec is posi ioned on he igh side o he page. To display he es o he in o ma- ion on each page, a block was used, called “con en _ igh side”, whose con en can be eplaced o ex ended in child empla es. To ende he in o ma ion abou he home page, ha sc ip was used and he images and espec i e cap ions, as well as he de ini ion o he day on which he con lic began, we e inse ed in he “con en _ igh side“ block. Fo he p ojec in oduc ion page, ano he empla e was c ea ed. This empla e ex ends he empla e “base.h ml”, he e o e i s con en will be combined in he base empla e. On his page, a sample o he p ojec ’s in oduc ion is p esen ed. On ano he page, he p ojec ’s 5.4. A isi o he web pla o m 53 Figu e 19: GUCO Websi e Co pus: Co pus Page A e submi ing he o m, he i les o he a ious esul ed a icles and/o pos s a e made a ailable on he websi e. Figu e 20 con ains he lis o esul s a e sea ching he co pus o an a icle in he newspape “Público”, which con ains he wo d “gue a” in i s i le and was published be ween he da es o Feb ua y 24,2022 and Sep embe 30,2023. Figu e 20: GUCO Websi e Co pus: Lis o Resul s Page To access each a icle o pos and espec i e commen s, he use mus choose one o he esul s. A e ha , all in o ma ion ela ing o ha a icle o pos is a ailable on a new web page. Figu es 21 and 22 p esen s he in o ma ion abou an a icle o he co pus. 5.4. A isi o he web pla o m 54 Figu e 21: GUCO Websi e Co pus: Resul Page (1) Figu e 22: GUCO Websi e Co pus: Resul Page (2) 6 CONCLUSION This Mas e ’s wo k aims o c ea e an online co pus ega ding he con lic be ween Russia and Uk aine ha began in Feb ua y 2022. In Chap e 1, he in oduc ion o he p oblem is desc ibed, ocusing on i s con ex ualiza ion and a b ie desc ip ion o ela ed p ojec s. This pa o he documen also explains he objec i es o his disse a ion, he me hodology app oach, he esea ch hypo hesis, and he s uc u e o he documen . Then, in Chap e 2, he s a e o he a is explo ed, whe e concep s such as co pus, CMC, and NLP a e add essed. Mo eo e , a ew examples o ela ed p ojec s a e lis ed and explained. A e wa d, a sys em a chi ec u e o he p ojec is sugges ed in Chap e 3. In his chap e , he chosen sou ces o in o ma ion a e e ealed and an explana ion o hei selec ion is p o ided. Addi ionally, ex ac o s c ea e o collec da a om he newspape “Público” we e also co e ed in his sec ion. Mo eo e , i is possible o see an example o a news s o y om he Po uguese online newspape “Público” as well as examples o he in o ma ion ex ac ed in JSON o ma . Chap e 4explains he co pus cons uc ion p ocess. This chap e p o ides a de ailed explana ion o how he ex ac o s we e c ea ed o collec in o ma ion om each sou ce. Some examples o ex ac ed in o ma ion sa ed in a JSON ile a e p o ided. Addi ionally, he gene al o ma in which he in o ma ion should be s o ed is desc ibed. Las ly, he chap e also explains he ex -cleaning p ocess made o he ex ac ed in o ma ion. A e ga he ing all he desi ed in o ma ion, he nex s ep was o s o e he in o ma ion in a eposi o y. In Chap e 5, he da abase used o s o e he in o ma ion is desc ibed. Then, a websi e was c ea ed so he use could explo e he co pus c ea ed. In his chap e , he p ocess o de eloping he websi e is p o ided. The ini ially designed mockups o he websi e a e p esen ed, along wi h explana ions o bo h he backend and on end componen s. Finally, a demons a ion o how o na iga e and u ilize he websi e is p o ided. The websi e can be accessed a : h ps://guco.epl.di.uminho.p /. 55 6.1. Fu u e Wo k 56 6.1 u u e wo k All he objec i es de ined a he beginning o his mas e p ojec we e comple ed. Howe e , he e a e se e al changes ha can be implemen ed o imp o e his wo k. In u u e wo k, he upda ing o he ex ac ed a icles and pos s is p oposed. In his wo k, he ex ac ion o he in o ma ion s opped a he end o July 2023. Howe e , he con lic pe sis s. The ex ac ion o a icles and pos s abou e en s a e he end o July 2023 should be done o keep he co pus up o da e. In he c ea ed websi e, he use can explo e he co pus and il e he in o ma ion using i s sou ce, he a icles o pos s’ da e, and using a lis o e ms o sea ch o he desi ed a icles o pos s compa ing hem o hei i le. As a p oposal o u u e wo k, i is sugges ed o ca ego ize he a icles and he pos s. In o he wo ds, i is sugges ed o di ide he in o ma ion in o a icles o pos s abou economics, pe sonal opinions, and mo e. Las ly, an addi ional ask is sugges ed. In his wo k, in he s age o ex ac ing in o ma ion om online newspape s, he p ocess is being ca ied ou and supe ised by a human. Tha is, a pe son has o choose among se e al a icles he ones ha bes i he con ex o his wo k and hen pass i s link as an a gumen o he ex ac o s c ea ed o ex ac he in o ma ion. A p oposal o u u e wo k would be o au oma e his s ep using a i icial in elligence app oaches. BIBLIOGRAPHY E ienne Bech , Leland McInnes, John Healy, Cha les-An oine Du e e, Immanuel WH Kwok, Lai Guan Ng, Flo en Ginhoux, and E an W Newell. Dimensionali y educ ion o isualizing single-cell da a using umap. Na u e bio echnology,37(1):38–44,2019. Cons ança Elias, Jo ge B andão Gonçal es, Ma ia A aújo, Ped o Pinhei o, C is iana A aújo, and Ped o Rangel Hen iques. Ne ac, an au oma ic classi ie o online ha e speech commen s. In Wo ld Con e ence on In o ma ion Sys ems and Technologies, pages 494–505. Sp inge , 2021. Manuel B. Ga cia and A mi Cunanan-Yabu . Public sen imen and emo ion analyses o wi e da a on he 2022 ussian in asion o uk aine. In 2022 9 h In e na ional Con e ence on In o ma ion Technology, Compu e , and Elec ical Enginee ing (ICITACEE), pages 242–247, 2022. doi: 10.1109/ICITACEE55701.2022.9924136. Hans WA Hanley, Deepak Kuma , and Zaki Du ume ic. Happens ance: U ilizing seman ic sea ch o ack ussian s a e media na a i es abou he usso-uk ainian wa on eddi . a Xi p ep in a Xi :2205.14484,2022. URL h ps://a xi .o g/pd /2205.14484.pd . I Hemala ha, GP Sa adhi Va ma, and A Go a dhan. P ep ocessing he in o mal ex o e icien sen imen analysis. In e na ional Jou nal o Eme ging T ends & Technology in Compu e Science (IJETTCS),1(2):58–61,2012. Ped o Rangel Hen iques, C is iana Es e es A aújo, Isabel E mida, and Idale e Dias. Sc aping news si es and social ne wo ks o p ejudice e m analysis. 2019. doi: h p:// eposi o ium.sdum.uminho.p /handle/1822/66512. Clay on Hu o and E ic Gilbe . Vade : A pa simonious ule-based model o sen imen analysis o social media ex . In P oceedings o he in e na ional AAAI con e ence on web and social media, olume 8, pages 216–225,2014. Moaiad Ahmad Khde . Web sc aping o web c awling: S a e o a , echniques, app oaches and applica ion. In e na ional Jou nal o Ad ances in So Compu ing and i s Applica ions, 13(3):144 –168,2021. URL h ps://www.scopus.com/inwa d/ eco d.u i?eid=2-s2. 0-85120853380&pa ne ID=40&md5=3b03613c6d348041161c889 859172b2. Ela Kuma . Na u al language p ocessing. IK In e na ional P L d, 2013. 57 BIBLIOGRAPHY 58 Quang Thai Le and Da a Pish a. Applica ion o web sc aping and google api se ice o op imize con enience s o es’ dis ibu ion. In 2015 17 h In e na ional Con e ence on Ad anced Communica ion Technology (ICACT), pages 478–482. IEEE, 2015. Leland McInnes, John Healy, and S e e As els. hdbscan: Hie a chical densi y based clus e - ing. J. Open Sou ce So w.,2(11):205,2017. Ryan Mi chell. Web sc aping wi h Py hon: Collec ing mo e da a om he mode n web. " O’Reilly Media, Inc.", 2018. A zucan Özgü , Le en Özgü , and Tunga Güngö . Tex ca ego iza ion wi h class-based and co pus-based keywo d selec ion. In In e na ional Symposium on Compu e and In o ma ion Sciences, pages 606–615. Sp inge , 2005. Shaimaa Fahad Rashid and Rawaa Pu os Qasha. Ex ac ing and a chi ing da a om social media o suppo cul u al he i age p ese a ion in nine eh. In 2022 In e na ional Con e ence on Compu e Science and So wa e Enginee ing (CSASE), pages 295–300,2022. doi: 10.1109/CSASE51777.2022.9759782. Alexande Romiszowski and Robin Mason. Compu e -media ed communica ion. In Handbook o esea ch on educa ional communica ions and echnology, pages 402–436. Rou ledge, 2013. Salim Sazzed. The dynamics o uk aine- ussian con lic h ough he lens o demog aphically di e se wi e da a. In 2022 IEEE In e na ional Con e ence on Big Da a (Big Da a), pages 6018–6024,2022. doi: 10.1109/BigDa a55660.2022.10020274. Alexande She so , Ch is os Tzagka akis, Despoina An onakaki, Poly ios P a ikakis, and So i is Ioannidis. Twi e da ase on he usso-uk ainian wa , 2022. Shi angi Singh, Reshma Rajan, Laasyas ee, D ishya Ramesh, Jasna Su esh, Vaishna i KP, and Anjali T. Co i igh : A co id-19 esou ce acke websi e wi h loca ion-wise sea ch using web sc aping. In 2022 4 h In e na ional Con e ence on Sma Sys ems and In en i e Technology (ICSSIT), pages 1326–1331,2022. doi: 10.1109/ICSSIT53264.2022.9716527. Uddagi i Si isha and Bolem Sai Chandana. Aspec based sen imen and emo ion analysis wi h obe a, ls m. In e na ional Jou nal o Ad anced Compu e Sci- ence and Applica ions,2022. doi: 10.14569/IJACSA.2022.0131189. URL h ps: //www.scopus.com/inwa d/ eco d.u i?eid=2-s2.0-85143858476&doi=10.14569% 2 IJACSA.2022.0131189&pa ne ID=40&md5= d26a29d43e0c603d533410dd 16c4bb . All Open Access, Gold Open Access. Da id Ma hew Thomas and Sandeep Ma hu . Da a analysis by web sc aping using py hon. In 2019 3 d In e na ional con e ence on Elec onics, Communica ion and Ae ospace Technology (ICECA), pages 450–454,2019. doi: 10.1109/ICECA.2019.8822022. BIBLIOGRAPHY 59 Yiming Zhu, Ehsan-ul Haq, Lik-Hang Lee, Ga e h Tyson, and Pan Hui. A eddi da ase o he usso-uk ainian con lic in 2022.a Xi p ep in a Xi :2206.05107,2022.