Enginee ing and Technology Jou nal e-ISSN: 2456-3358
Volume 10 Issue 10 Oc obe -2025, Page No.-7279-7284
DOI: 10.47191/e j/ 10i10.08, I.F. – 8.482
© 2025, ETJ
7279
ETJ Volume 10 Issue 10 Oc obe 2025, Si i Monalisa
F om Sen imen o S a egy: Emo ional Classi ica ion o Cus ome
Engagemen in Co po a e Cosme ic Video Ma ke ing
Si i Monalisa*
Depa emen o In o ma ion Sys ems, Facul y o Science and Technology, Uni e si as Islam Nege i Sul an Sya i Kasim,
Pekanba u-Riau, Indonesia.
ABSTRACT: This s udy examines a ious emo ions exp essed in cus ome engagemen con en a ailable in co po a e p oduc
ideos on he TikTok pla o m. The esea ch iden i ied 301 pos s wi h a o al o 39,615 commen da a collec ed om he o icial
TikTok accoun . The da a unde wen a p ep ocessing s age o ensu e accu a e p ocessing, ollowed by s emming, classi ica ion
using TF-IDF, Bag-o -Wo ds isualiza ion, and classi ica ion wi h he NRC EmoLex lib a y o de ec emo ions p esen in he
da ase . The indings e eal ha , in addi ion o posi i e and nega i e commen s, a ious emo ions such as Ange , An icipa ion,
Disgus , Fea , Joy, Sadness, Su p ise, T us , Posi i e, and Nega i e we e iden i ied. The highes emo ion sco e was Posi i e (33%),
ollowed by Joy (17%), An icipa ion (16%), and Su p ise (12%). These esul s indica e ha he con en analyzed in his s udy has a
subs an ial posi i e impac on he audience, wi h posi i e emo ions domina ing. Howe e , companies should emain a en i e o he
p esence o An icipa ion and Nega i e emo ions. The e o e, i is essen ial o main ain ac o s ha igge posi i e emo ions,
minimize con en ha elici s nega i e emo ions, and manage sensi i e con en ca e ully o a oid gene a ing ad e se esponses.
KEYWORDS: Bag-o -Wo ds (BoW), EmoLex Lib a y, Emo ions, Nega i e Sen imen , Posi i e Sen imen , TikTok, and Video
Commen s
I.
INTRODUCTION
The eme gence o a ious communica ion media oday has
encou aged companies o le e age mul iple channels o
in e ac wi h cus ome s [1]. The a ailabili y o channels such
as websi es, social media, and ideo-sha ing pla o ms has
become an essen ial componen o co po a e ma ke ing
s a egies. Th ough hese channels, use s can di ec ly access
ideo con en wi hou downloading i [2] and a e also able o
commen , eply, and exp ess likes o dislikes on sha ed ideo
con en [3].
Cu en ly, bo h companies and cus ome s u ilize social
media o in e ac wi h ideo c ea o s o o he iewe s.
Companies o en ake ad an age o ideo commen s o
highligh hei p oduc ad an ages, while cus ome s use
social media o sha e hei expe iences and seek opinions and
eedback om o he use s be o e making a pu chase decision.
Howe e , no all ideo con en has he same e ec i eness in
engaging cus ome s o in luencing consume s’ pu chase
in en ions. Se e al elemen s in luence cus ome s’ emo ional
engagemen , including e bal na a ion, acial exp essions,
and one o oice in ideos. The e o e, companies mus pay
a en ion o hese elemen s when de eloping con en
s a egies. Mo eo e , use commen s and in e ac ions on
ideos e lec he le el o engagemen and pe cep ion owa d
he p oduc . Al hough con en s a egy plays a c ucial ole in
ma ke ing, he e a e s ill limi a ions in unde s anding how
ideo con en elemen s a ec cus ome engagemen and
pu chase in en ion.
The inc easing amoun o cus ome engagemen con en
a ailable in co po a e p oduc ideos may con ain hidden
dimensions and deepe emo ions ha ul ima ely in luence
cus ome s’ pu chasing decisions [1]. Emo ional engagemen
d i es deepe in e ac ions, whe e cus ome s who expe ience
speci ic emo ions end o commen , like, sha e ideo con en ,
and a e expec ed o build an emo ional bond wi h he b and
and os e long- e m loyal y [4]. In addi ion, speci ic
emo ions exe a s onge in luence on decisions han gene al
posi i e a i udes. The e o e, his s udy in es iga es how
a ious emo ions a e exp essed in cus ome engagemen
con en ea u ed in co po a e p oduc ideos.
In he con ex o Indonesia, he cosme ic indus y has
expe ienced signi ican g ow h o e he pas decade, d i en
by inc easing consume awa eness o beau y ends, sel -ca e,
and social media in luence [5]. Da a om he Indonesian
Minis y o Indus y indica es ha he cosme ic sec o is
among he as es -g owing indus ies in he coun y,
suppo ed by a la ge millennial and Gen Z popula ion ha
ac i ely engages wi h digi al pla o ms o beau y inspi a ion
and p oduc ecommenda ions [6]. The ise o e-comme ce
pla o ms and sho - ideo applica ions such as TikTok has
u he accele a ed his end, p o iding b ands wi h
oppo uni ies o p omo e p oduc s h ough c ea i e and
in e ac i e con en [7]. Consequen ly, beau y and cosme ic
b ands in Indonesia ha e adop ed ideo-based ma ke ing
s a egies o a ac and engage cus ome s, le e aging
in luence s and b and-gene a ed con en o s eng hen b and
“F om Sen imen o S a egy: Emo ional Classi ica ion o Cus ome Engagemen in Co po a e Cosme ic Video
Ma ke ing”
7280
ETJ Volume 10 Issue 10 Oc obe 2025, Si i Monalisa
loyal y and d i e pu chase in en ions.
P oduc ideos gene ally o igina e om wo main ypes:
hose c ea ed by companies hemsel es, known as b and-
gene a ed con en , and hose p oduced by consume s using
he p oduc o se ice, known as use -gene a ed con en . This
s udy ocuses on b and-gene a ed ideos on he TikTok
pla o m. Videos p oduced by he company a e used o
p edic cus ome beha io al ou comes. Cus ome beha io
can be measu ed h ough ideo commen s by analyzing
commen s, eplies, likes, and o he o ms o in e ac ion,
which ul ima ely lead o posi i e and nega i e
ecommenda ions in luencing consume s’ pu chasing
decisions [8].
P e ious li e a u e has examined how cus ome s espond
o hese ideos h ough commen ing, liking o disliking, and
eplying [9], [10]. All hese ac ions e lec he g owing ole
o cus ome engagemen on online ideo pla o ms, which
ul ima ely assis s p ospec i e buye s in making pu chasing
decisions [11] and, in u n, de e mines he success o ailu e
o a company’s p oduc o b and. Pos - ideo cus ome
beha io s ha e also inc eased in he o m o likes o dislikes,
commen s, and eplies in a ious con ex s [3], [12], [13].
Wi h he ise o social media usage, such beha io s ha e
become mo e easily obse able by cus ome s An inc ease in
social media engagemen has enhanced he isibili y o hese
beha io s o cus ome s [1], [14].
II.
RESEARCH METHOD
This s udy collec ed cus ome engagemen con en om
p oduc ideos a ailable on he TikTok pla o m, ocusing on
all commen s in hei o iginal o m. Commen s om selec ed
ideos we e ex ac ed and o ganized in o a s uc u ed da ase
in abula o ma , consis ing o quali a i e ex en ies.
The p ep ocessing phase in ol ed se e al s eps o ensu e
da a quali y. Ini ially, a cleaning p ocess was conduc ed o
emo e elemen s such as emo icons, hash ags, men ions,
excessi e spaces, numbe s, duplica e en ies, and commen s
wi h ewe han i e wo ds [15]. Subsequen ly, case olding
was applied o con e all ex o lowe case, ollowed by
okeniza ion o spli sen ences in o indi idual wo ds,
no maliza ion o s anda dize linguis ic o ms, and s op-wo d
emo al o elimina e non-in o ma i e e ms. S emming was
hen pe o med o educe wo ds o hei oo o m. These
p ocedu es acili a ed he emo al o i ele an e ms and
enabled he iden i ica ion o signi ican Bag-o -Wo ds (BoW)
ea u es using he Te m F equency–In e se Documen
F equency (TF-IDF) algo i hm.
The sen imen labeling s age employed a lexicon-based
app oach using Py hon o classi y wo ds as Posi i e o
Nega i e. To ensu e accu acy, sen imen assignmen s we e
alida ed by an independen expe . Addi ionally, TF-IDF
was u ilized o quan i y e m signi icance by e alua ing i s
dis ibu ion ac oss documen s [16]. The BoW ep esen a ion
was hen isualized h ough a wo d cloud o depic he mos
equen e ms.
Finally, emo ion classi ica ion was pe o med using he
NRC Wo d-Emo ion Associa ion Lexicon (EmoLex) h ough
he R p og amming language. This esou ce acili a ed he
ex ac ion o posi i e and nega i e sen imen s, along wi h
eigh p ima y emo ions: joy, us , su p ise, an icipa ion,
ange , ea , sadness, and disgus [16]. Emo ion sco es we e
compu ed based on he equency o e ms associa ed wi h
each emo ional ca ego y. P io s udies ha e ex ensi ely
documen ed he applicabili y o EmoLex-based sen imen
analysis ac oss a ious esea ch domains [17], [18], [19].
III.
RESULT AND DISCUSSION
The da ase o his s udy comp ised 301 TikTok pos s
collec ed h ough a web sc aping app oach, yielding a o al o
39,615 commen s om he o icial TikTok accoun .
Following an ini ial inspec ion, a comp ehensi e da a
cleaning p ocess was conduc ed o ensu e accu acy and
consis ency. A e cleaning, he da ase was educed o 23,341
commen s. The cleaning p ocess was implemen ed in wo
s ages: a manual e iew o emo e i ele an en ies and
au oma ed il e ing using Py hon sc ip s o de ec and
elimina e noisy da a, such as emp y alues o unusable
en ies.
The ex p ep ocessing s age in ol ed mul iple sequen ial
s eps. Fi s , case olding was pe o med o con e all ex in o
lowe case o ma , ensu ing uni o mi y ac oss he da ase .
This was ollowed by okeniza ion, in which sen ences we e
segmen ed in o indi idual okens. Tokeniza ion was ca ied
ou using he Na u al Language Toolki (NLTK) lib a y in
Py hon, which is widely ecognized o ex p ocessing asks.
Nex , da a no maliza ion was applied o co ec non-
s anda d language o ms and abb e ia ions commonly ound
in use -gene a ed con en on social media pla o ms. Fo
example, e ms such as “uda” we e no malized o “sudah”
(al eady), “yg” o “yang” (which), and “sblm” o “sebelum”
(be o e). Despi e hese e o s, he no maliza ion p ocess did
no ully s anda dize all in o mal a ia ions, lea ing some
inconsis encies in he da ase . The ou comes o okeniza ion
and no maliza ion a e p esen ed in Tables 1 and 2.
Following no maliza ion, s opwo d emo al was
conduc ed o elimina e wo ds ha lack seman ic ele ance o
do no con ibu e meaning ully o he analysis. This s ep
aimed o imp o e he quali y o he ex ep esen a ion by
disca ding high- equency bu low-in o ma ion e ms.
Examples o emo ed s opwo ds include “kakk,” “ini” ( his),
“li e,” “sudah” (al eady), “sebelum” (be o e), “ idak” (no ),
“a au” (o ), “mau” (wan ), “nya,” and “yang” (which). By
emo ing hese e ms, he da ase was e ined o include only
linguis ically signi ican okens, acili a ing mo e accu a e
subsequen analyses.
These p ep ocessing s eps collec i ely ensu ed ha he
da ase was p epa ed o ad anced ex mining echniques,
such as Bag-o -Wo ds (BoW) modeling, Te m F equency–
“F om Sen imen o S a egy: Emo ional Classi ica ion o Cus ome Engagemen in Co po a e Cosme ic Video
Ma ke ing”
7281
ETJ Volume 10 Issue 10 Oc obe 2025, Si i Monalisa
In e se Documen F equency (TF-IDF) weigh ing, sen imen
labeling, and emo ion classi ica ion, which a e discussed in
he ollowing sec ions.
Table 1. Sample O Tokenized Da a
N
o
Tokenizing
1
[“kakk”, “li e”, “kapann”, “uda”, “nungguin”,
“ini”]
2
[“dh”, “ au”, “komenku”, “sblum”, “li e”]
3
[“yg”, “mau”, “beli”, “scinca e”, “cammile”,
“ba u”, “pakai”, “diki ”, “dimu ahin”]
4
[“aaa”, “ke inggalan”, “li e”, “camille”, “habis”,
“ke idu an”]
5
[“kak”, “mau”, “ anya”, “maske ”, “camille”,
“yang”, “ a ian”, “apa”, “ya”, “un uk”, “komedo”,
“mendem”]
...
...
23.34
1
[“nan i”, “mlm, “ad”, “ap”,”ini”, “kaka”]
Table 1. Sample O No malized Da a
N
o
Tokenizing
1
[“kakk”, “li e”, “kapan”, “sudah”, “nungguin”,
“ini”]
2
[“sudah”, “ ahu”, “komenku”, “sebelum”, “li e”]
3
[“yang”, “mau”, “beli”, “skinca e”, “cammile”,
“ba u”, “pakai”, “diki ”, “dimu ahin”]
4
[“aaa”, “ke inggalan”, “li e”, “camille”, “habis”,
“ke idu an”]
5
[“kak”, “mau”, “ anya”, “maske ”, “camille”,
“yang”, “ a ian”, “apa”, “ya”, “un uk”, “komedo”,
“ e sumba ”]
...
...
23.34
1
[“nan i”, “mlm, “ada”, “apa”,”nih”, “kaka”]
The inal s age o ex p ep ocessing was s emming, which
se es o educe wo d a ia ions by con e ing okens o hei
oo o ms, he eby simpli ying he ex ep esen a ion wi hin
he aining da a. This p ocess was implemen ed using he
Sas awi module in Py hon. Fo ins ance, wo ds such as
“masknya” we e con e ed o “mask”, “komenku” o
“komen”, “nungguin” o “ unggu”, and “p oduknya” o
“p oduk”.
Following ex p ep ocessing, sen imen labeling was
conduc ed using a lexicon-based app oach implemen ed in
Py hon. In his s age, each commen was assigned ei he a
posi i e o nega i e label based on i s sen imen sco e.
Howe e , he labeling p ocess e ealed some commen s wi h
null alues o classi ied as neu al sen imen . To add ess his,
manual cleaning was pe o med o emo e commen s wi hou
posi i e o nega i e sen imen alues.
Subsequen ly, he numbe o commen s in each sen imen
ca ego y was calcula ed. The esul s indica ed ha 14,869
commen s we e classi ied as posi i e ( ep esen ed in g een),
whe eas 8,472 commen s we e classi ied as nega i e
( ep esen ed in ed). Figu e 4.2 illus a es he p opo ion o
sen imen analysis esul s o he TikTok commen da ase o
Camille Beau y O icial. Based on he isualiza ion, he
highes p opo ion co esponds o posi i e sen imen a
63.7% (g een), while nega i e sen imen accoun s o 36.3%
( ed).
The classi ica ion esul s indica e ha he numbe o
posi i e commen s subs an ially exceeds he numbe o
nega i e commen s, sugges ing ha , o e all, use esponses
owa d he analyzed TikTok con en we e p edominan ly
posi i e. None heless, he p esence o a signi ican p opo ion
o nega i e commen s highligh s he need o con inued
a en ion.
Following he classi ica ion s age, TF-IDF was applied o
compu e e m weigh s and assign nume ical alues o he
ex ac ed e ms. The ea e , a Bag-o -Wo ds (BoW)
app oach was u ilized, isualized in he o m o a wo d cloud,
whe e he mos equen ly occu ing wo ds appea in la ge
on sizes. A subse o commen s was andomly selec ed om
he o iginal da ase o his isualiza ion. The analysis
e ealed dominan e ms such as “camille”, “bange ”,
“dape ”, “beli”, and “kak”, which a e signi ican indica o s o
pu chase in en . The BoW isualiza ion is p esen ed in
Figu e 1.
Figu e 1. Bag-o -Wo ds Visualiza ion o Posi i e
Sen imen
In he TikTok commen s o Camille Beau y, he wo mos
equen ly occu ing wo ds associa ed wi h posi i e sen imen
(indica ed in g een) we e “ hank” and “bi hday.” In con as ,
o nega i e sen imen (indica ed in ed), he mos equen
wo ds we e “pe ih” and “kalah.”
The nex s age in ol ed emo ion analysis, which was
pe o med in Google Colabo a o y using EmoLex (NRC
Wo d-Emo ion Associa ion Lexicon) o ex ac posi i e and
nega i e sen imen s as well as associa ed emo ions, including
Ange , An icipa ion, Disgus , Fea , Joy, Sadness, Su p ise,
and T us (Mohammad & Tu ney, 2012). Posi i e sen imen
was associa ed wi h he emo ions An icipa ion, Joy, Su p ise,
T us , and Posi i e, while nega i e sen imen was linked o
Ange , Disgus , Fea , Sadness, and Nega i e.
Figu e 2 p esen s he esul s o he emo ion classi ica ion,
showing ha he ca ego ies Posi i e and Joy eco ded he
“F om Sen imen o S a egy: Emo ional Classi ica ion o Cus ome Engagemen in Co po a e Cosme ic Video
Ma ke ing”
7282
ETJ Volume 10 Issue 10 Oc obe 2025, Si i Monalisa
highes sco es, ollowed by T us and An icipa ion, which
we e also ela i ely high. This indica es ha he da a
p edominan ly con eys op imism and us . Con e sely,
nega i e ca ego ies such as Ange , Disgus , Fea , and Sadness
had lowe sco es and we e he e o e no dominan . The
emo ions Su p ise and Nega i e we e also p esen bu no
signi ican .
The emo ion analysis e ealed en ca ego ies: Ange ,
An icipa ion, Disgus , Fea , Joy, Sadness, Su p ise, T us ,
Posi i e, and Nega i e (Figu e 3). The highes -sco ing
emo ion was Posi i e (33%), ollowed by Joy (17%),
An icipa ion (16%), and Su p ise (12%).
These indings indica e ha Camille Beau y’s con en has a
p edominan ly posi i e impac on i s audience, wi h posi i e
emo ions being dominan . Howe e , he p esence o
An icipa ion and Nega i e emo ions sugges s ha he
company should con inue o os e posi i e igge s, educe
con en ha e okes nega i e esponses, and ca e ully manage
sensi i e ma e ial o p e en ad e se eac ions.
Figu e 2. Emo ion Analysis Resul s Using R
Figu e 3. Emo ional Sco e Su ey Resul s
Based on Figu e 3, posi i e emo ions domina e, wi h he
highes sco es eco ded in he Posi i e ca ego y a 32.7% and
Joy a 16.6%. This indica es ha he majo i y o TikTok use s
esponded posi i ely o he con en and p oduc s o Camille
Beau y O icial, sugges ing ha he b and main ains a
a o able image in he eyes o i s cus ome s. An icipa ion
(15.5%) and Su p ise (12.3%) u he sugges cu iosi y and
in e es owa d new p oduc s o p omo ional o e s. The T us
sco e o 7.0% e lec s a deg ee o cus ome con idence in he
b and, al hough he e is oom o imp o emen h ough
enhanced c edibili y and anspa ency.
Con e sely, nega i e emo ions we e also p esen , albei in
smalle p opo ions. The Nega i e ca ego y sco ed 6.9%, wi h
Fea accoun ing o 3.7%, Disgus 2.1%, and Ange 1.5%.
These indings indica e he p esence o dissa is ac ion ela ed
o ce ain aspec s such as p oduc quali y, p icing, o se ice.
Sadness egis e ed only 1.5%, signi ying ha deeply nega i e
complain s we e a e.
The analysis sugges s ha Camille Beau y O icial has
subs an ial po en ial o s eng hen cus ome loyal y by
le e aging posi i e emo ions such as Joy, An icipa ion, and
Su p ise h ough c ea i e campaigns, in e ac i e con en , and
he in oduc ion o appealing new p oduc s. Howe e ,
a en ion mus also be gi en o nega i e emo ions o p e en
a decline in cus ome sa is ac ion. Recommended s a egies
include imp o ing cus ome se ice quali y, ensu ing p oduc
consis ency, and p o iding clea communica ion on issues
ha may p o oke Fea o Disgus . By ein o cing T us and
minimizing nega i e pe cep ions, Camille Beau y’s b and
image can become mo e posi i e and compe i i e in he
ma ke . Table 3 p esen s ecommenda ions and co ec i e
s a egies aimed a enhancing cus ome sa is ac ion and
s eng hening he b and image o Camille Beau y O icial.
Table 3. P oposed Recommenda ions And S a egic
Imp o emen s Fo The Resea ch Objec
Recommenda ion /
S a egy
Associa ed Emo ions
1
Op imize en e aining and
educa ional con en
signi ican ly o enhance
engagemen [20]
Joy (17%) and
An icipa ion (16%)
domina e among he
audience, indica ing
high in e es and
cu iosi y owa d
engaging con en .
2
Inc ease posi i e e iews as
hey enhance us and os e
loyal y among bo h new and
exis ing cus ome s [21]
T us has a ela i ely
high sco e bu equi es
ein o cemen o
s eng hen loyal y.
Posi i e emo ions such
as Posi i e (33%), Joy
(17%), and
An icipa ion (16%) a e
dominan .
3
Implemen in luence
ma ke ing o inc ease
An icipa ion (16%) and
Su p ise (12%) e lec
“F om Sen imen o S a egy: Emo ional Classi ica ion o Cus ome Engagemen in Co po a e Cosme ic Video
Ma ke ing”
7283
ETJ Volume 10 Issue 10 Oc obe 2025, Si i Monalisa
engagemen , ollowe s, and
iewe s by eaching a
b oade audience on social
media [22]
en husiasm owa d
in e ac i e ac i i ies.
4
Enhance cus ome
expe ience and use
in e ac ion as hese
in luence elec onic wo d-
o -mou h (eWOM), which is
essen ial o imp o ing
b and isibili y and us
[21]
Nega i e emo ions
such as Fea (3.7%)
and Disgus (2.1%)
should be add essed,
while T us (7%) needs
o be s eng hened.
IV.
CONCLUSION AND IMPLICATIONS
This s udy in es iga ed cus ome engagemen on TikTok
p oduc ideos o Camille Beau y O icial, ocusing on
sen imen pola i y and emo ional classi ica ion de i ed om
use -gene a ed commen s. Th ough sys ema ic ex
p ep ocessing, sen imen labeling using a lexicon-based
app oach, and emo ion analysis u ilizing he NRC EmoLex
lexicon, he esea ch iden i ied dominan sen imen pa e ns
and associa ed emo ional s a es.
The indings indica e ha posi i e sen imen s domina e
(63.7%), signi ican ly su passing nega i e sen imen s
(36.3%), sugges ing an o e all a o able pe cep ion o
Camille Beau y’s TikTok con en . Emo ion classi ica ion
u he e ealed ha Posi i e (33%), Joy (17%), and
An icipa ion (16%) we e he mos p ominen emo ional
ca ego ies, e lec ing op imism, en husiasm, and cu iosi y
owa d he b and’s o e ings. Addi ionally, Su p ise (12%)
and T us (7%) demons a e subs an ial audience engagemen
and b and con idence. In con as , nega i e emo ions such as
Fea (3.7%), Disgus (2.1%), and Ange (1.5%)— hough less
p e alen —signal a eas o po en ial dissa is ac ion ela ed o
p oduc quali y, p icing, o se ice expe iences.
V.
LIMITATIONS AND FUTURE RESEARCH
Despi e i s con ibu ions, his s udy has limi a ions ha
wa an conside a ion. Fi s , he analysis was con ined o
TikTok commen s on a single b and’s accoun , which may
es ic gene alizabili y ac oss pla o ms o indus ies.
Second, he eliance on lexicon-based sen imen and emo ion
classi ica ion may o e look con ex -speci ic nuances such as
sa casm o mixed emo ions. Fu u e esea ch could adop
ad anced na u al language p ocessing (NLP) echniques,
such as deep lea ning models (e.g., BERT o LSTM), o
imp o e classi ica ion accu acy and con ex ual
unde s anding. Addi ionally, compa a i e s udies ac oss
mul iple social media pla o ms o b ands could p o ide
b oade insigh s in o emo ional engagemen pa e ns.
Inco po a ing beha io al me ics such as click- h ough a es,
pu chase in en ions, and con e sion da a may also enhance
he p edic i e alue o sen imen -emo ion analysis.
REFERENCES
1. S. R. Ag awal, “Digi al Pollu ion and I s Impac on
he Family and Social In e ac ions,” J. Fam. Issues,
ol. 42, no. 11, pp. 2648–2678, No . 2021, doi:
10.1177/0192513X20985558.
2. Aus e be y Da id, “The Technology o Video and
Audio S eaming Second Edi ion,” 2005.
3. Q. Gaus, A. Jolli , and M. A. Mo eno, “A con en
analysis o YouTube dep ession pe sonal accoun
ideos and hei commen s,” Compu . Hum. Beha .
Rep., ol. 3, Jan. 2021,
doi: 10.1016/j.chb .2020.100050.
4. G. C. Wicaksana and E. A dyan, “Emo ional
Expe iences D i e Cus ome Loyal y in Indonesia,”
Acad. Open, ol. 9, no. 2, July 2024, doi:
10.21070/acopen.9.2024.9144.
5. N. Nawiyah, R. C. Kaemong, M. A. Ilham, and F.
Muhammad, “PENYEBAB PENGARUHNYA
PERTUMBUHAN PASAR INDONESIA
TERHADAP PRODUK SKIN CARE LOKAL
PADA TAHUN 2022,” ARMADA J. Peneli .
Mul idisiplin, ol. 1, no. 12, pp. 1390–1396, Dec.
2023, doi: 10.55681/a mada. 1i12.1060.
6. R. Sus aning um, “PURCHASE INTENTION
GENERASI Z PADA PRODUK KOSMETIK
DENGAN TEKNOLOGI AUGMENTED
REALITY PADA MASA PANDEMI,” J. Ilm.
Ekon. Bisnis, ol. 28, no. 3, pp. 405–419, 2023, doi:
10.35760/eb.2023. 28i3.7170.
7. E. S. C. Nangoy, J. R. E. Tampi, and T. M. Tumbel,
“Peman aa an Aplikasi Tik ok sebagai Digi al
Ma ke ing P omo ion pada Cu abeau y Manado,”
P oduc i i y, ol. 5, no. 2, pp. 859–863, June 2024,
doi: 10.35797/ejp. 5i1.54704.
8. R. Fauzan Abdillah and A. N. P ames i, “SEMINAR
NASIONAL AMIKOM SURAKARTA
(SEMNASA) 2024 DAMPAK RATING DAN
ULASAN KONSUMEN TERHADAP
KEPUTUSAN PEMBELIAN DI E-COMMERCE”.
9. D. Röche , G. K. Shahi, G. Neubaum, B. Ross, and
S. S iegli z, “The Ne wo ked Con ex o COVID-19
Misin o ma ion: In o ma ional Homogenei y on
YouTube a he Beginning o he Pandemic,” Online
Soc. Ne w. Media, ol. 26, No . 2021,
doi: 10.1016/j.osnem.2021.100164.
10. W. Ta esse, “YouTube ma ke ing: how ma ke e s’
ideo op imiza ion p ac ices in luence ideo iews,”
In e ne Res., ol. 30, no. 6, pp. 1689–1707, Oc .
2020, doi: 10.1108/INTR-10-2019-0406.
11. A. Pansa i and V. Kuma , “Cus ome engagemen :
he cons uc , an eceden s, and consequences,” J.
Acad. Ma k. Sci., ol. 45, no. 3, pp. 294–311, May
2017, doi: 10.1007/s11747-016-0485-6.
12. L. Dessa and V. Pi a di, “How s o ies gene a e
“F om Sen imen o S a egy: Emo ional Classi ica ion o Cus ome Engagemen in Co po a e Cosme ic Video
Ma ke ing”
7284
ETJ Volume 10 Issue 10 Oc obe 2025, Si i Monalisa
consume engagemen : An explo a o y s udy,” J.
Bus. Res., ol. 104, pp. 183–195, No . 2019, doi:
10.1016/j.jbus es.2019.06.045.
13. S. Moldo an, Y. S einha , and D. R. Lehmann,
“P opaga o s, C ea i i y, and In o ma i eness:
Wha Helps Ads Go Vi al,” J. In e ac . Ma k., ol.
47, pp. 102–114, Aug. 2019,
doi: 10.1016/j.in ma .2019.02.004.
14. A. N. Smi h, E. Fische , and C. Yongjian, “How
Does B and- ela ed Use -gene a ed Con en Di e
ac oss YouTube, Facebook, and Twi e ?,” J.
In e ac . Ma k., ol. 26, no. 2, pp. 102–113, May
2012, doi: 10.1016/j.in ma .2012.01.002.
15. S. R. Ag awal and D. Mi al, “Op imizing cus ome
engagemen con en s a egy in e ail and E- ail:
A ailable on online p oduc e iew ideos,” J.
Re ail. Consum. Se ., ol. 67, July 2022,
doi: 10.1016/j.j e conse .2022.102966.
16. S. M. Mohammad and P. D. Tu ney,
“CROWDSOURCING A WORD-EMOTION
ASSOCIATION LEXICON,” 2012. [Online].
A ailable:
h p://c owdsou cing. ypepad.com/cs/2006/06
17. S. Cha e jee, “Explaining cus ome a ings and
ecommenda ions by combining quali a i e and
quan i a i e use gene a ed con en s,” Decis.
Suppo Sys ., ol. 119, pp. 14–22, Ap . 2019, doi:
10.1016/j.dss.2019.02.008.
18. Y. Dang, Y. Zhang, and H. Chen, “A Lexicon-
Enhanced Me hod o Sen imen Classi ica ion: An
Expe imen on Online P oduc Re iews,” 2010.
[Online]. A ailable: www.compu e .o g/in elligen
19. M. Taboada, J. B ooke, M. To iloski, K. Voll, and
M. S ede, “Lexicon-Based Me hods o Sen imen
Analysis,” 2011.
20. S. Feb iyan i and A. Budiman, “Pendampingan
Peningka an K ea i i as Kon en Digi al Sebagai
Upaya Meningka kan Engagemen Media Sosial
Pada 86 P oduc ion”.
21. A is Ku niawan, Lili Ma linah, Yosie No e ha, and
Vina Islami, “Penga uh Digi al Ma ke ing, Social
Media Engagemen , dan Cus ome T us Te hadap
Loyali as Pelanggan E-Comme ce,” El-Mal J. Kaji.
Ekon. Bisnis Islam, ol. 5, no. 10, Oc . 2024, doi:
10.47467/elmal. 5i10.5268.
22. F. Wa dah, “Analisis Penga uh In luence e hadap
Mina Beli Konsumen pada Pe usahaan Ja aMi i,”
ol. 02, no. 03, 2023.
