Vol.:(0123456789)
1 3
Social Ne wo k Analysis and Mining (2023) 13:29
h ps://doi.o g/10.1007/s13278-023-01033-8
ORIGINAL ARTICLE
Nuclea ene gy: Twi e da a mining o social lis ening analysis
Ena aZa abei ia‑Bilbao1· Mai eJaca‑Mada iaga1· RosaMa íaRio‑Bel e 2· IzaskunÁl a ez‑Meaza1
Recei ed: 8 Decembe 2022 / Re ised: 18 Janua y 2023 / Accep ed: 20 Janua y 2023
© The Au ho (s) 2023
Abs ac
Knowing he p esence, a i ude and sen imen o socie y is impo an o p omo e policies and ac ions ha in luence he
de elopmen o di e en ene gy sou ces and e en mo e so in he case o an ene gy sou ce such as nuclea , which has no been
wi hou con o e sy in ecen yea s. The pu pose o his pape was o conduc a social lis ening analysis o nuclea ene gy
using Twi e da a mining. A o al o 3,709,417 global wee s we e analyzed h ough he in e ac ions and emo ions o Twi e
use s h oughou a c ucial yea : 6mon hs be o e and 6mon hs a e he beginning o Russian in asion o Uk aine and he
i s a ack on he Zapo izhzhia NPP. The esea ch uses a no el app oach o combine social ne wo k analysis me hods wi h
he applica ion o a i icial neu al ne wo k models. The esul s e eal he digi al con e sa ion is in luenced by he Russian
in asion o Uk aine. Howe e , wee s con aining pe sonal opinions o in luen ial people also manage o en e he digi al
con e sa ion, de ining he magni ude and di ec ion o he deba e. The digi al con e sa ion is no cons uc ed as a public
a gumen . Gene ally, i is a con e sa ion wi h non-pola ized communi ies (poli ics, business, science and media); nei he
a med con lic o mili a y h ea s agains Zapo izhzhia NPP succeed in ousing an i-nuclea oices, e en hough hese e en s
do modi y he o ien a ion o he sen imen in he language used, making i mo e nega i e.
Keywo ds Nuclea ene gy· Twi e · Social ne wo k analysis· A i icial neu al ne wo ks· Russia–Uk aine con lic
1 In oduc ion
I is an undeniable ac ha in ecen yea s he e has been an
in ense wo ldwide public deba e on nuclea ene gy (Diaz-
Mau in 2014; Diaz-Mau in and Ko acic 2015). The u gency
o mi iga ing clima e change, as well as dependence on ossil
uels, has eopened he deba e on his ene gy sou ce (F ied-
e ich and Boud y 2022). Thus, while i s de ac o s call o i s
aboli ion due o i s dange ousness and he p oblems de i ed
om was e gene a ion (G eenpeace 2022; NRDC 2022), i s
de ende s undamen ally see i as a way owa d deca boniza-
ion (IAEA 2020; NEI 2022).
The Uni ed S a es (his o ical suppo e o his ene gy
sou ce and coun y wi h mos nuclea powe plan s in
ope a ion (IAEA-PRIS 2022) is cau iously emb acing
nuclea powe , despi e ce ain en i onmen alis s’ pe sis en
conce ns, o help achie e i s goal o a ne -ze o ca bon econ-
omy o he na ion by 2050 (USDS-USEOP 2021). Japan
(coun y ope a ing he hi d mos nuclea powe plan s),
whe e nuclea powe has been ana hema since he Fukush-
ima Daiichi nuclea powe plan disas e in 2011 (Schneide
and F ogga 2012), is ad oca ing he ope a ion o eac o s
ha ha e emained inac i e o en yea s since he acciden
(Reu e s 2021).
The e is no doub ha nuclea ene gy- ela ed de elop-
men s ha e also been compulsi e in Eu ope. In No embe
2021, F ance (coun y ope a ing he second mos nuclea
powe plan s) announced ha i will elaunch he cons uc-
ion o nuclea eac o s in he coun y (LeFiga o 2021),
hus e e sing he discussion on an ene gy sou ce ha
seemed obsole e a e he ca as ophe a he Fukushima
plan . In addi ion, on New Yea ’s E e 2021, nuclea powe
expe ienced a pa ial shu down in Be lin and a esu gence
in B ussels: Ge many shu down h ee o i s six ope a -
ing nuclea powe plan s as pa o he app o ed plan
o shu down all a omic p oduc ion by he end o 2022
(Joly 2021). Howe e , a he same s oke o midnigh , he
Eu opean Commission announced i s pionee ing p oposal
o change he classi ica ion o g een ene gy o include
* Ena a Za abei ia-Bilbao
ena a.za [email p o ec ed]
1 Facul y o Enginee ing, Uni e si y o heBasque Coun y
(UPV/EHU, Bilbao, Spain
2 Facul y o Enginee ing, Uni e si y o heBasque Coun y
(UPV/EHU, Vi o ia-Gas eiz, Spain
Social Ne wo k Analysis and Mining (2023) 13:29
1 3
29 Page 2 o 17
nuclea powe and na u al gas as g een ene gy (Eu opean
Commission 2022).
In his scena io, howe e , an e en ha has changed
he wo ldwide deba e abou nuclea ene gy has been he
Russian in asion o Uk aine on Feb ua y 24, 2022. The
a med con lic be ween Russia and Uk aine has undoub -
edly in luenced ene gy policies a ound he wo ld and
has led coun ies o push o a apid ansi ion o g eene
ene gy sou ces. Nuclea powe has been in decline since
he Fukushima disas e mo e han a decade ago, bu he
ene gy and en i onmen al c isis is p omp ing eapp aisal
o he indus y. Hence, he Russian in asion o Uk aine;
ising na u al gas p ices; and Russian ene gy dependence
ha e eopened he deba e on whe he nuclea powe can
help sol e he challenges o ene gy secu i y and clima e
change.
None heless, he Ma ch 4, 2022, a ack o he Zapo -
izhzhia nuclea powe plan (NPP) in Uk aine ( he la ges
nuclea powe plan in Eu ope) has caused eno mous social
uphea al. The a ack on he Zapo izhzhia NPP is he i s
a ack, wi h pe nicious in en , on a nuclea powe plan
(wi hou a adia ion acciden ). The e ha e been p e ious
nuclea disas e s such as Kysh ym (USSR, 1957), Windscale
Piles (UK, 1957), Th ee Mile Island (USA, 1979), Che no-
byl (USSR, 1986) and Fukushima (Japan, 2011); howe e ,
none we e in en ionally p o oked.
Knowing he p esence, a i ude and sen imen o socie y
is impo an o p omo e policies and ac ions ha in luence
he de elopmen o di e en ene gy sou ces (Iba -Alonso
e al. 2022). Hence, being awa e ha in he pos -Fukushima
e a, nuclea ene gy became a delica e si ua ion and public
opinion became pessimis ic (Kim e al. 2016), so analyzing
how ecen e en s ha e a ec ed socie y can be conside ed
impo an and o g ea in e es o he scien i ic communi y.
Tha is o say, wha he social esponse on his occasion
has been and whe he i has led o any change in s a egy o
posi ioning wi h espec o his ene gy sou ce.
To his end, he in e ac ions and discussions gene a ed
on Twi e will be analyzed. Twi e is conside ed a social
ne wo k ha ep esen s an ideal scena io o di e se in e -
ac i e audiences and a con empo a y digi al own squa e
(Fe nández-Gómez e al. 2018; O begozo-Te adillos e al.
2022).
Based on his scena io, he pu pose o his pape was o
conduc a social lis ening analysis o nuclea ene gy. Wi h
his pu pose, on he one hand, he ela ionships es ablished
(discussion communi ies gene a ed) on he social ne wo k
in ela ion o nuclea ene gy will be analyzed h oughou a
whole yea and; on he o he hand, he o e all eelings and
a i udes owa d nuclea ene gy will be de e mined h ough
sen imen analysis o wee s. All his, wi h special emphasis
on he Russian in asion o Uk aine and he Zapo izhzhia
NPP a ack.
2 Twi e da a mining andnuclea ene gy
The in o ma ion age o in o ma ion socie y is de ined by
he social changes ha ha e aken place since he inal
decades o he wen ie h cen u y, de i ed om he de el-
opmen o In o ma ion and Communica ion Technologies
oge he wi h he de elopmen o ne wo ked social s uc-
u es, which ha e impac ed all a eas o human ac i i y
(Mansell 2009; Cas ells 2010; Del-F esno-Ga cía 2014).
Today, a dense social ne wo k o in e ac ions links people,
in o ma ion, e en s, places, e c., acili a ing o limi ing he
lows o in o ma ion, ideas and pe cep ions, among o he s,
in an ins an aneous and massi e ne wo ked communica-
ion sys ems (Del-F esno-Ga cía 2014).
In his con ex , Twi e is a discu si e space ha has
eme ged in he hea o he digi al ago a, allowing people
o make hemsel es isible: being seen and hea d unde
he p o ec ion o he mos di e se causes, slogans o si ua-
ions o con o mi y o noncon o mi y (Bae 2016). Se e al
s udies ha e shown how Twi e plays a key ole in imes
o c isis and con lic (O begozo-Te adillos e al. 2019).
In addi ion, i makes i possible o cap u e a wide a ie y
o da a in eal ime, and e ospec i ely, p o iding access
o da a eco ds o human ac i i y in he digi al sphe e o e
ime (Del-F esno-Ga cía 2014).
None heless, he alue o he da a alone is no high;
howe e , da a a e he aw ma e ial om which mo e alue
can be gene a ed. The speci ic p ocess used o ha ness da a
as a aw ma e ial and gene a e is known as da a mining,
and he ope a ion essen ially consis s o cap u ing a se ies
o in o ma ion eco ds and in e p e ing hem (Mo ales-i-
G as 2020a). In his ega d, nume ous esea che s ha e
used Twi e da a o achie e di e en objec i es ela ed o
nuclea ene gy h ough he analysis o such da a and using
a ious da a mining echniques.
In his scena io, i is wo h no ing ha mos o he sci-
en i ic wo ks ela ed o Twi e da a mining and nuclea
ene gy e e o he disas e a he Fukushima nuclea
powe plan . As a consequence o he Fukushima nuclea
ca as ophe, la ge amoun s o adioac i e ma e ials leaked
ou , causing adioac i e pollu ion o wa e . Hence, sho ly
a e he ca as ophe, public opinion was o med h ough
a ious pla o ms, including digi al social ne wo k se -
ices such as Twi e (Seung-Hoi e al. 2016).
The exchange o in o ma ion on digi al social ne wo ks
has a - eaching posi i e e ec s, such as eal ime and high
b oadcas abili y p ope ies. Mo eo e , Twi e use s a e
simul aneously consume s and con ibu o s o in o ma-
ion (Veil e al. 2011). Howe e , in he Fukushima powe
plan disas e , he in o ma ion ha sp ead apidly included
misleading epo s, such as he claim ha iodine is use ul
o ea ing adioac i i y as a subs i u e o s able iodine,
Social Ne wo k Analysis and Mining (2023) 13:29
1 3
Page 3 o 17 29
and ha became a p oblem (Aoki e al. 2018). Risk in o -
ma ion ansmi ed by digi al social ne wo ks had a s ong
in luence on bo h he pe cep ion o isk and public eac-
ions o an unusual ca as ophe (Chung 2018). All in all,
a e he Fukushima nuclea powe plan acciden , isk
and en i onmen al communica ion has become a widely
s udied opic h ough he social ne wo k Twi e (Li e al.
2016). In he mon hs ollowing he acciden , science-based
wee s dec eased; howe e , wee s wi h mo e emo ional
exp essions began o sp ead and Twi e wo ked as a new
public sphe e especially in e ms o an i-nuclea mo e-
men s (Kim 2014). Likewise, i was obse ed ha he
ole o in luence s was c ucial o sp eading in o ma ion
(Tsuboku a e al. 2018; Sano e al. 2021).
In his con ex , wi h he eme gence o social ne wo ks,
new ways o exp essing opinions ha e eme ged, making he
analysis o emo ions and eelings displayed on digi al pla -
o ms key owa d be e unde s anding use s' opinions on a
gi en opic (A umugam e al. 2021). Hence, Twi e use s’
emo ional esponse a e he Fukushima nuclea disas e ,
mainly nega i e and opposed o nuclea ene gy, has been
s udied in de ail (Miu a e al. 2015; Su e al. 2016; Kim
e al. 2016; Hasegawa e al. 2020). Iden i ying he p ocess
by which people emo ionally espond could be use ul in
isk communica ion when simila disas e s occu in u u e
(Hasegawa e al. 2020).
Howe e , despi e he ac ha he Fukushima inciden has
been widely analyzed h ough Twi e da a mining, o he
ields o s udy as well as o he geog aphical loca ions ha e
also been co e ed, such as poli ical deba es abou nuclea
wi hd awal ini ia i es o na a i e policy amewo ks (A l
e al. 2018; Gup a e al. 2018).
In all hese s udies, a ious da a mining ools, echniques
and me hodologies a e used. One o he echniques widely
used is social ne wo k analysis (SNA) (Ran asila e al. 2018;
Yagaha a e al. 2018), which allows knowledge o be gene -
a ed acco ding o he ela ional s uc u e o i ual in e -
ac ions (Mo ales-i-G as 2020b). In addi ion, inc easingly
sophis ica ed ools a e being used o p o ide mo e p ecise
answe s o mo e complex eali ies. In his sense, wi hin deep
lea ning algo i hms, a i icial neu al ne wo ks (ANNs) a e
used, especially o deciphe ing public opinion o analyzing
emo ions and sen imen s owa d nuclea ene gy (Liu and Na
2018; Kha ua e al. 2020; A umugam e al. 2021). Howe e ,
hese echniques a e usually used sepa a ely, i.e., SNA ech-
niques a e no combined wi h ANN.
3 Resea ch ques ions andme hodology
The s udy ocuses on he in e ac ions and emo ions o Twi -
e use s ega ding nuclea ene gy h oughou a c ucial yea :
6mon hs be o e and 6mon hs a e he beginning o Russian
in asion o Uk aine and he i s a ack on he Zapo izhzhia
NPP, i.e., om Sep embe 1, 2021, o Augus 31, 2022. Th ee
speci ic pe iods ha e been s udied: Fi s pe iod, be o e he
beginning o he con lic ; second pe iod, he i s h ee weeks
o he con lic , when he i s a ack on he nuclea powe plan
ook place; and hi d pe iod, a e he beginning o he con lic ,
i.e., a e he i s h ee weeks ha e elapsed.
Fo ha pu pose, he esea ch uses a no el app oach o
combine social ne wo k analysis (SNA) me hods wi h he
applica ion o a i icial neu al ne wo k (ANN) models. In
o he wo ds, a se o me hods o he analysis o social in e -
ac ions ha speci ically in es iga e ela ional s uc u es and
hei ep esen a ion as ne wo ks (SNA) a e combined wi h
algo i hms modeled as elemen a y uni s o neu ons con-
nec ed in such a way o o m a ne wo k capable o sol ing
complex nonlinea p oblems (ANN) (see Analysis me hods
sec ion).
3.1 Resea ch ques ions
In o de o add ess he objec i e o social lis ening analysis
on Twi e abou nuclea ene gy using Twi e da a mining,
he ollowing ques ions ha e been used as a guide:
RQ-1: Wha ha e been he main cha ac e is ics o he
o e all con e sa ion abou nuclea ene gy ha has aken
place h oughou a yea ? A e he e mo phological di e -
ences be ween he con e sa ions gene a ed be o e, du -
ing and a e he beginning o he Russian in asion o
Uk aine and he a ack on he Zapo izhzhia NPP?
RQ-2: Which ha e been he main communi ies gene a ed
in he o e all con e sa ion abou nuclea ene gy h ough-
ou a yea ? How ha e hese communi ies changed be o e,
du ing and a e he beginning o he Russian in asion o
Uk aine and he a ack on he Zapo izhzhia NPP?
RQ-3: Who ha e been he leade s in he o e all con e sa-
ion abou nuclea ene gy h oughou a yea ? How ha e
hese in luen ial playe s changed be o e, du ing and a e
he beginning o he Russian in asion o Uk aine and he
a ack on he Zapo izhzhia NPP?
RQ-4: Wha has been he sen imen abou nuclea powe
h oughou a yea ? How has he sen imen exp essed in
nuclea ene gy- ela ed wee s changed be o e, du ing and
a e he beginning o he Russian in asion o Uk aine
and he a ack on he Zapo izhzhia NPP?
3.2 Resea ch me hodology
3.2.1 Collec ing ex da a o analysis
The i s s ep, in o de o add ess he main objec i e o he
s udy, was da a ex ac ion o he Twi e social ne wo k and
da a p epa a ion.
Social Ne wo k Analysis and Mining (2023) 13:29
1 3
29 Page 4 o 17
In his ega d, i s , he sea ch que y conside ed o selec
he wee s ha make up he sample o his s udy was:
“nuclea ene gy” OR nuclea ene gy OR “nuclea powe ” OR
nuclea powe . Subsequen ly, he wee s ( om Sep embe 1,
2021, o Augus 31, 2022) we e downloaded using he Twi -
e API o Academic Resea ch (Twi e 2022) and Twa c
(a command line ool and Py hon lib a y o collec ing and
a chi ing Twi e JSON da a ia he Twi e API (Twa c
2022)). The da a we e collec ed e oac i ely in wo ba ches.
On Augus 25, 2022, wee s sen om Sep embe 1, 2021,
h ough July 31, 2022, we e downloaded, and on Sep embe
1, wee s sen h oughou Augus 2022 we e downloaded.
Once he wee s we e ob ained, da a p ep ocessing was
necessa y o subsequen da a analysis. Thus, OpenRe ine
open-sou ce so wa e was used o da a cleanup and ans-
o ma ion (OpenRe ine 2022). This cleaning and ans o -
ma ion consis ed, mainly, in ex ac ing men ions in o de o
syn hesize di e en ne wo ks based on which use s men ion
which o he use s in he con e sa ion i sel .
3.2.2 Analysis me hods
The second s ep was he empi ical app oach o he esea ch,
de eloped in wo s ages. In he i s s age, ne wo ks me ics
and main communi ies we e analyzed (answe ing RQ-1,
RQ-2 and RQ-3), and in he second s age, he emo ional
s age o he digi al discussions was s udied (answe ing
RQ-4).
To ob ain he di e en discussion ne wo ks gene a ed
abou nuclea ene gy (a ne wo k ha akes in o accoun he
discussion o he whole yea , ano he ne wo k o he i s
pe iod, ano he o he second pe iod and he las one o he
hi d pe iod), all he men ions in he digi al con e sa ions
we e ex ac ed, syn hesizing ne wo ks based on which use s
men ioned o he use s in he con e sa ion i sel . The esul -
ing ne wo ks we e expo ed o he social ne wo k analysis
open-sou ce so wa e Pajek (Ba agelj and M a 2002), and
in o de o analyze he es ablished ela ionships be ween
use s o he digi al con e sa ion sphe e, di e en me ics
we e ob ained, a bo h he global and indi idual le el (a
ne wo k le el and a nodal le el). Hence, h ough he global
me ics, he mo phology o he ne wo ks was analyzed, and
h ough he me ics a he nodal le el, he key ac o s o he
con e sa ions we e iden i ied.
The Lou ain mul i-le el algo i hm was applied o iden-
i y he communi ies ha ha e pa icipa ed in he digi al
con e sa ions. This algo i hm allows densely in e connec ed
g oups o nodes (Twi e use ) o be gene a ed, as well as
ob aining he bes ne wo k pa i ions o he highes modula -
i y igu e. Thus, he g oups o clus e s ep esen a i e o he
digi al con e sa ion o he di e en pe iods unde analysis
we e iden i ied.
The syn hesized ne wo ks and communi ies de ec ed in
he p e ious s eps we e u he p ocessed wi h Gephi so -
wa e, an open-sou ce ool o p o en alidi y o imp o ing
he isualiza ion and analysis o la ge ne wo k g aphs (Bas-
ian e al. 2009; O begozo-Te adillos e al. 2022). Gephi’s
Fo ce-A las 2 algo i hm (Jacomy e al. 2014) was employed
because o i s use ulness in b inging nodes ha o m pa o
he same communi ies close oge he and away om hose
wi h which hey a e less algo i hmically ela ed.
In he second s age o he empi ical app oach, a i icial
neu al ne wo ks we e used o p ocess he wee s and achie e
he goal o analyzing he sen imen s gene a ed in hose
wee s. Sen imen analysis, also known as opinion mining o
emo ion classi ica ion, is a combina ion o na u al language
p ocessing (NLP) and ex mining, wi h he aim o analyzing
he ex da a o social media, among o he s, and ha ing ha
in o ma ion, mine he use ’s emo ions (Salloum e al. 2017;
Jain and Kaushal 2018).
I has been de ec ed ha a i icial neu al ne wo ks o ally
su pass mo e adi ional machine lea ning models (Jain
and Kaushal 2018). Mo eo e , he key o achie e a high
accu acy model seems o be in c ea ing an a chi ec u e ha
combines di e en deep neu al ne wo ks (Sosa 2017; Kamiş
and Goula as 2019; Ume e al. 2021). The e o e, his s udy
used a model adap ed om Pe iwal (2021) ha p oduces a
sen imen analysis o wee s using deep lea ning algo i hms.
Thus, he a chi ec u e o he neu al ne wo k model is led by
he Bi-LSTM and a en ion as de ailed below.
To do ha , i s o all, unp ocessed ex da a om he
sen imen 140 da ase we e ga he ed and p ep ocessed (Go
e al. 2009; Kaggle 2022; Sen imen 140 2022). This da a-
se con ains 1,600,000 wee s ex ac ed using he Twi e
API, which ha e al eady had hei sen imen classi ied as
a posi i e, labeled as 1, o nega i e, labeled as 0. The e
a e 800,000 wee s o each classi ica ion ype espec-
i ely, which means ha he da ase is no skewed. A e
p ep ocessing, da a we e spli in o a aining da ase and
a es da ase . Then, ex was okenized and padded. The
ou pu achie ed is he inpu o ou sen imen classi ica ion
neu al ne wo k model, which is a deep lea ning sequence
model. The model has an a chi ec u e composed o he ol-
lowing laye s: an embedding laye c ea ed using he wo d-
2 ec model o con e wee s in o wo d ec o s (Mikolo
e al. 2013); a Bi-LSTM neu al ne wo k laye o cap u e
he seman ic meaning o he ex (El aik and N aoui 2021;
Chand a e al. 2021); an a en ion mechanism laye o ex ac
he mos ele an wo ds; a dense laye which adds a ully
connec ed laye in he model and ha he a gumen passed
speci ies he amoun o nodes in ha laye ; and inally, a las
dense laye wi h a sigmoid ac i a ion unc ion o achie e he
emo ion classi ica ion (Fig.1). A e aining and es ing he
desc ibed model, i was applied o unlabeled wee s ela ed
o nuclea ene gy, e u ning he emo ion o each o hem.
Social Ne wo k Analysis and Mining (2023) 13:29
1 3
Page 5 o 17 29
Figu e1 summa izes he used sen imen analysis neu al ne -
wo k model (Kaggle 2022).
Figu e2 summa izes he me hodological p ocess applied
in he s udy.
4 Resul s anddiscussion
The esea ch collec ed a o al sample o 3,858,024 wee s.
Since English e ms we e used o he sea ch, almos all he
wee s we e in English (96.15%), so in o de o main ain
homogenei y h oughou he s udy, i was decided o wo k
wi h he wee s in English, i.e., wi h a sample o 3,709,417
wee s.
Figu e3 shows ha , as expec ed, he dis ibu ion o
wee s has no been homogeneous o e ime. Speci ically,
unusual ac i i y is obse ed wi h 606,055 wee s (16.3% o
o al wee s) on Ma ch 4, 2022. The a ack on Zapo izhzhia,
Eu ope’s la ges nuclea powe plan (CNN 2022), gene a ed
global social ala m e lec ed in he social ne wo k ac i i y.
Rega ding he h ee pe iods unde s udy, 18.5% o he
wee s collec ed co espond o pe iod I, 35.7% o he wee s
co espond o pe iod II and 45.8% o he wee s co espond
o pe iod III (Fig.3). Thus, he impac ha he con lic has
had on he in ensi ica ion o he digi al con e sa ion abou
nuclea ene gy can be obse ed.
By isola ing he h ee pe iods o a clea e iew o wha
happens in each o he pe iods, Figs.4, 5 and 6 show ha he
dis ibu ion o wee s has no been homogeneous o e ime.
Di e en p o agonis s, opics o e en s ha e led o inc eased
use ac i i y on he social ne wo k.
The i s pe iod (Fig.4) shows odd ac i i y on Decembe
8 and 9, due o a wee pos ed by Elon Musk suppo ing
nuclea powe plan s: “Unless suscep ible o ex eme na u al
disas e s, nuclea powe plan s should no be shu down.”1
Likewise, on Feb ua y 13, an Ame ican al - igh poli ical
ac i is , Jack Posobiec, pos ed a wee ela ed o Ame ican
poli ics, which had a g ea esponse: “Biden’s Na ional
Fig. 1 Sen imen analysis neu al ne wo k model applied in he s udy
Fig. 2 Me hodological p ocess
applied in he s udy
Iden i y main que y e minology
(“nuclea ene gy” o nuclea ene gy o
“nuclea powe ” o nuclea powe ) and
imespan (09/01/2021-08/31/2022)
Twi e da a
ex ac ion & p epa a ion
Collec da a (Twi e API o
Academic Resea ch and Twa c)
Da a cleaning and p epa a ion
(OpenRe ine)
Da a analysis & isualiza ion
wi h big da a echniques
Ne wo ks me ics and main
communi y analysis (Pajek, Gephi,
Powe Que y)
Sen imen analysis (Deep lea ning
algo i hms: model using wo d2 ec, Bi-
LSTM and a en ion mechanism)
Achie emen o objec i e /
conclusions
Social Ne wo k Analysis
A i icial Neu onal Ne wo ks
1 h ps:// wi e . com/ elonm usk/ s a us/ 14686 89628 16105 2674.
Social Ne wo k Analysis and Mining (2023) 13:29
1 3
29 Page 6 o 17
Secu i y Ad iso was in ol ed in a illegal spying ope a ion
on he p e ious P esiden . He is now escala ing ensions
wi h a nuclea powe in Eu asia bc Biden’s poll numbe s
a e down.”2 In addi ion, on Feb ua y 22, he p oximi y o
he ou b eak o he Russian in asion is no ed on he ne wo k.
The second pe iod (Fig.5), as p e iously emphasized,
shows he impac o he Russian in asion o Uk aine on
ne wo k ac i i y ela ed o nuclea ene gy and he g ea
social uphea al gene a ed by he a ack on he Zapo izhzhia
NPP.
In he las pe iod (Fig.6), he unusual ne wo k ac i i y
on July 6 is due o he Eu opean Pa liamen 's accep ance o
he inclusion o nuclea ene gy and gas as en i onmen ally
sus ainable economic ac i i ies (Eu opean Pa liamen 2022).
Finally, he up u n in he numbe o wee s in Augus is,
abo e all, due o he upsu ge o con inuous Russian a acks
o he Zapo izhzhia NPP and wee s in a o o nuclea
ene gy pos ed by in luen ial ac o s such as Elon Musk:
0
100000
200000
300000
400000
500000
600000
1-Sep
5-Sep
9-Sep
13-Sep
17-Sep
21-Sep
25-Sep
29-Sep
3-Oc
7-Oc
11-Oc
15-Oc
19-Oc
23-Oc
27-Oc
31-Oc
4-No
8-No
12-No
16-No
20-No
24-No
28-No
2-Dec
6-Dec
10-Dec
14-Dec
18-Dec
22-Dec
26-Dec
30-Dec
3-Jan
7-Jan
11-Jan
15-Jan
19-Jan
23-Jan
27-Jan
31-Jan
4-Feb
8-Feb
12-Feb
16-Feb
20-Feb
24-Feb
28-Feb
4-Ma
8-Ma
12-Ma
16-Ma
20-Ma
24-Ma
28-Ma
1-Ap
5-Ap
9-Ap
13-Ap
17-Ap
21-Ap
25-Ap
29-Ap
3-May
7-May
11-May
15-May
19-May
23-May
27-May
31-May
4-Jun
8-Jun
12-Jun
16-Jun
20-Jun
24-Jun
28-Jun
2-Jul
6-Jul
10-Jul
14-Jul
18-Jul
22-Jul
26-Jul
30-Jul
3-Aug
7-Aug
11-Aug
15-Aug
19-Aug
23-Aug
27-Aug
31-Aug
Nº o wee s
Da e (yea s 2021-2022)
Pe iod I
Pe iod II
Pe iod III
Fig. 3 Daily end o wee s h oughou he yea ( om Sep embe 1, 2021, o Augus 31, 2022)
0
2000
4000
6000
8000
10000
12000
14000
Nº o wee s
Da e (pe iod I)
Twee pos ed by Elon Musk
suppo ing nuclea powe plan s Twee pos ed by Jack Posobiec
ela ed o Ame ican poli ics
Days p io o he
Russian in asion
Fig. 4 Daily end o wee s o e he i s pe iod ( om Sep embe 1, 2021, o Feb ua y 23, 2022)
Fig. 5 Daily end o wee s
o e he second pe iod ( om
Feb ua y 24, 2022, o Ma ch
16, 2022)
0
100000
200000
300000
400000
500000
600000
700000
Nº o wee s
Da e (pe iod II)
A ack on he
Zapo izhzhia NPP
2 h ps:// wi e . com/ jackp osobi ec/ s a us/ 14928 74434 72574 4643.
Social Ne wo k Analysis and Mining (2023) 13:29
1 3
Page 7 o 17 29
0
5000
10000
15000
20000
25000
30000
35000
40000
45000
Nº o wee s
Da e (pe iod III)
Eu opean Pa liamen 's
accep ance o nuclea ene gy as
en i onmen ally sus ainable
economic ac i i y
Con inuous a acks o he Zapo izhzhia
NPP and Twee pos ed by Elon Musk in
a o o nuclea ene gy
Fig. 6 Daily end o wee s o e he hi d pe iod ( om Ma ch 17, 2022, o Augus 31, 2022)
3 h ps:// wi e . com/ elonm usk/ s a us/ 15632 92201 04343 1424.
Table 1 Me ics ex ac ed om
he analyzed con e sa ions
(Za abei ia-Bilbao e al. 2022b)
a Numbe o use s who ha e pa icipa ed in he con e sa ion, i.e., only use s who ha e in e pella ed o ha e
been in e pella ed.
Me ics analyzed O e all ne wo ks o di e en pe iods
Global Pe iod I Pe iod II Pe iod III
To al impac s ( wee s o e wee s) 3,709,417 684,961 1,325,015 1,699,441
Use s (nodes)a1,243,554 350,377 574,937 668,543
A e age impac s (pe use ) 2.98 1.95 2.3 2.54
A cs (in e ac ions) 3,767,699 758,625 1,436,134 1,697,983
Densi y 0.00000244 0.00000618 0.00000434 0.00000380
A e age deg ee 6.05956637 4.33033561 4.99579606 5.07965232
Maximum dis ance 28 24 28 32
A e age dis ance 7.27924 6.64913 10.04885 7.61276
Inpu deg ee cen aliza ion 0.07459838 0.05662457 0.07531547 0.06023185
Ou pu deg ee cen aliza ion 0.00236256 0.00425212 0.00143234 0.00362949
Be weenness cen aliza ion 0.01229997 0.00685957 0.00209164 0.00790441
Numbe o clus e s 20,985 12,186 9,326 13,388
Modula i y (Lou ain mul i-le el algo i hm) 0.688192 0.760213 0.665745 0.717920
“Coun ies should be inc easing nuclea powe gene a ion!
I is insane om a na ional secu i y s andpoin & bad o
he en i onmen o shu hem down.”3
All in all, i is obse ed ha he ole o in luence s such
as Elon Musk is c ucial o sp eading messages.
4.1 Ne wo ks me ics andmain communi y analysis
(RQ‑1, RQ‑2, RQ‑3)
4.1.1 Ne wo k me ics
Me ics ex ac ed om he analyzed con e sa ions show
he main cha ac e is ics o he con e sa ions abou nuclea
ene gy in he di e en pe iods and mo phological di e -
ences be ween hem (Table1).
As men ioned in he p e ious sec ion, he numbe o
impac s in he second pe iod (1,325,015 impac s in 3weeks)
was ela i ely la ge (35.7% o he impac s o he whole
yea ). This was somewha o eseeable due o he media and
social commo ion gene a ed by he Russian in asion o
Uk aine and he a ack on he Zapo izhzhia NPP.
I can be seen ha he mos cohesi e ne wo k is ha o he
i s pe iod (densi y 0.000618%); howe e , he low-densi y
igu es indica e ha he e a e s ill many s a egic in e pel-
la ions o be explo ed by he nuclea ene gy ecosys em.
Mo eo e , inpu deg ee cen aliza ion is highe in he case
o he second pe iod (7.53%) han o he o he pe iods. The
g oup o nodes ha ecei e many men ions is la ge han
in he es o he cases, placing us in a scena io o a less
ho izon al dialogic. Howe e , i is obse ed ha all cen al-
i y alues a e low. Thus, i is in e ed, bo h o he di e en
pe iods and o he o e all se , ha small g oups o ac o s
Social Ne wo k Analysis and Mining (2023) 13:29
1 3
29 Page 8 o 17
did no capi alize he ecep ion o men ions in he digi al
discussions (inpu deg ee cen aliza ion); ha he e was no
single g oup ha pos ed mos o he men ions (ou pu deg ee
cen aliza ion); and ha he digi al ne wo k is dis ibu ed
ho izon ally a he han being monopolized by a ew use s
(be weenness cen aliza ion).
Finally, in all cases, he modula i y is g ea e han 0.3,
which indica es, o he whole ime span s udied, a commu-
ni y s uc u e o g ea ma hema ical signi icance (Newman
and Gi an 2004).
All in all, excep o he high ela i e ac i i y in he sec-
ond pe iod, he e a e no signi ican mo phological di e -
ences in he ne wo ks o he di e en ime spans analyzed.
4.1.2 Main communi y analysis
Rega ding he cen al cha ac e s o he digi al discussion,
Figs.7, 8, 9 and 10 illus a e, pe di e en ime span s ud-
ied, which eigh main communi ies ( ep esen ing in all cases
mo e han 3.4% o he ac o s in he ne wo k) ha e pa ici-
pa ed in he digi al con e sa ion and he posi ion ha each
communi y holds.4
Tables2, 3, 4 and 5 show who he key playe s o he con-
e sa ion and wha he main opics discussed in each o he
communi ies ha e been. On he one hand, he analysis o he
p o agonis s o he con e sa ions has been conduc ed by ana-
lyzing bo h he unweigh ed and weigh ed inpu deg ee. The
i s o hese me ics indica es he numbe o “neighbo s” a
use has wi hou coun ing he numbe o imes he use alks
o he “neighbo s”, i.e., which use s ha e he la ges numbe
o unique audiences. Howe e , his me ic is combined wi h
he weigh ed inpu deg ee, since o long s udy pe iods, i
indica es ha a use is impo an no only in a conjunc u al
way, ha ing only once been e wee ed a lo , bu also in a
s uc u al way. And, on he o he hand, he con en ha has
been impo an in each o he communi ies has been de e -
mined by ca e ully eading he mos “ i alized” wee s in
each communi y.
Fi s o all, i can be seen ha he e is alignmen be ween
he agen s wi h he highes inpu deg ee and he highes
weigh ed inpu deg ee. In addi ion, no accoun displays, a
p io i, suspicious beha io . No use has a suspiciously high
weigh ed inpu deg ee le el compa ed wi h he unweigh ed
inpu deg ee le el, so ha , a p io i, no accoun is a i icially
p omo ed by o he au oma ed accoun s.
F om he communi y and leade analysis ca ied ou , i
can be in e ed ha , in gene al, ac o s ela ed o poli ics,
business, science and media lead he digi al discussions
abou nuclea ene gy on Twi e . The e o e, his social ne -
wo k is a dissemina ion ool ha wo ks, in his case, as a
means o exp ession o speci ic communi ies: Namely ech-
nical o collegia e communi ies, wi h a knowledge o in e es
in he opic, and ha suppo nuclea ene gy as a na ional,
in e na ional and en i onmen al s a egy.
Fig. 7 Mos impo an communi ies’ ne wo k h oughou he yea ( om Sep embe 1, 2021, o Augus 31, 2022)
4 The o de o he colo s will be main ained o all he ne wo ks and
will be in acco dance wi h he size o he communi y ( om la ges
o smalles ). The e o e, he ac ha a colo is epea ed om one ne -
wo k o ano he does no mean ha i is he same communi y, in ac ,
om one ne wo k o ano he he e will always be di e ences in he
communi ies.
Social Ne wo k Analysis and Mining (2023) 13:29
1 3
Page 9 o 17 29
Fig. 8 Mos impo an communi ies’ ne wo k o e he i s pe iod ( om Sep embe 1, 2021, o Feb ua y 23, 2022)
Fig. 9 Mos impo an communi ies’ ne wo k o e he second pe iod ( om Feb ua y 24, 2022, o Ma ch 16, 2022)
Social Ne wo k Analysis and Mining (2023) 13:29
1 3
29 Page 16 o 17
Ba agelj V, M a A (2002) Pajek—analysis and isualiza ion o la ge
ne wo ks. Sp inge , Be lin, Heidelbe g, pp 477–478
Cas ells M (2010) The Rise o he Ne wo k Socie y, 2nd edn.
Chand a S, Kuma Gou isa ia M, Gm H, e al (2021) Seman ic analysis
o sen imen s h ough web-mined wi e co pus. In: In e na ional
seman ic in elligence con e ence,. Delhi, pp 25–27
Chung IJ (2018) In e ac i e dynamics o media communica ion, isk
pe cep ion and public esponse o Fukushima nuclea disas e .
Ko ea Assoc Policy S ud 27:281–303
CNN (2022) Ma ch 4, 2022 Russia-Uk aine news
Eu opean Commission (2022) EU Taxonomy: Commission begins
expe consul a ions
Del-F esno-Ga cía M (2014) Haciendo isible lo in isible: isuali-
zación de la es uc u a de las elaciones en ed en Twi e po
medio del análisis de edes sociales. P o La In 23:246–252.
h ps:// doi. o g/ 10. 3145/ epi. 2014. may. 04
Diaz-Mau in F (2014) Going beyond he nuclea con o e sy. En i-
on Sci Technol 48:25–26. h ps:// doi. o g/ 10. 1021/ ES405 282Z/
ASSET/ IMAGES/ LARGE/ ES- 2013- 05282Z_ 0001. JPEG
Diaz-Mau in F, Ko acic Z (2015) The un esol ed con o e sy o e
nuclea powe : a new app oach om complexi y heo y. Glob
En i on Chang 31:207–216. h ps:// doi. o g/ 10. 1016/j. gloen cha.
2015. 01. 014
El aik H, N aoui E-H (2021) Deep a en ional bidi ec ional LSTM o
A abic sen imen analysis in Twi e . In: 2021 1s In e na ional
con e ence on eme ging sma echnologies and applica ions
(eSma TA), pp 1–8
Fe nández-Gómez JD, He nández-San aolalla V, Sanz-Ma cos P
(2018) In luence s, pe sonal b anding and poli ical ideology on
Twi e . Cuade nos.in o. h ps:// doi. o g/ 10. 7764/ cdi. 42. 1348
F iede ich S, Boud y M (2022) E hics o nuclea ene gy in imes o
clima e change: escaping he collec i e ac ion p oblem. Philos
Technol 35:1–27. h ps:// doi. o g/ 10. 1007/ s13347- 022- 00527-1
Go A, Bhayani R, Huang L (2009) Twi e sen imen classi ica ion
using dis an supe ision. CS224N P ojec Rep 1:2009
González V (2011) Educación pa a la ciudadanía democ á ica en la
cul u a digi al. Comumica 18:131–138. h ps:// doi. o g/ 10. 3916/
C36- 2011- 03-0
G eenpeace (2022) #Nuclea . h ps:// www. g een peace. o g/ in e na io
nal/ ag/ nucle a /. Accessed 6 Oc 2022
Gup a K, Ripbe ge J, Wehde W (2018) Ad ocacy g oup messaging
on social media: using he na a i e policy amewo k o s udy
Twi e messages abou nuclea ene gy policy in he Uni ed S a es.
Policy S ud J 46:119–136. h ps:// doi. o g/ 10. 1111/ psj. 12176
Hasegawa S, Suzuki T, Yagaha a A e al (2020) Changing emo ions
abou ukushima ela ed o he ukushima nuclea powe s a ion
acciden —how umo s de e mined people’s a i udes: social
media sen imen analysis. J Med In e ne Res 22:18662. h ps://
doi. o g/ 10. 2196/ 18662
IAEA (2020) D i ing deepe deca boniza ion wi h nuclea ene gy.
h ps:// www. iaea. o g/ bulle in/ d i i ng- deepe - deca boniz a ion-
wi h- nucle a - ene gy. Accessed 6 Oc 2022
IAEA-PRIS (2022) The Da abase on Nuclea Powe Reac o s. h ps://
p is. iaea. o g/ p is/ home. aspx. Accessed 8 Oc 2022
Iba -Alonso R, Qui oga-Ga cía R, A enas-Pa a M (2022) Opinion
mining o g een ene gy sen imen : a Russia–Uk aine con lic
analysis. Ma hema ics 10:1–22. h ps:// doi. o g/ 10. 3390/ ma h1
01425 32
Jacomy M, Ven u ini T, Heymann S, Bas ian M (2014) Fo ceA las2, a
con inuous g aph layou algo i hm o handy ne wo k isualiza-
ion designed o he Gephi so wa e. PLoS One 9:e98679. h ps://
doi. o g/ 10. 1371/ jou n al. pone. 00986 79
Jain K, Kaushal S (2018) A compa a i e s udy o machine lea ning
and deep lea ning echniques o sen imen analysis. In: 2018
7 h In e na ional con e ence on eliabili y, in ocom echnologies
and op imiza ion (T ends and Fu u e Di ec ions) (ICRITO), pp
493–487
Joly J (2021) Ge many begins nuclea phase-ou , shu s down h ee o
six nuclea powe plan s . eu onews
Kaggle (2022) Sen imen 140 da ase wi h 1.6 million wee s. h ps://
www. kaggle. com/ da as e s/ kazan o a/ sen i men 1 40. Accessed 4
Oc 2022
Kamiş S, Goula as D (2019) E alua ion o deep lea ning echniques
in sen imen analysis om Twi e da a. In: P oc—2019 In Con
Deep Lea n Mach Lea n Eme g Appl Deep 2019 12–17. h ps://
doi. o g/ 10. 1109/ DEEP- ML. 2019. 00011
Kha ua A, Camb ia E, Ho S, Na JC (2020) Deciphe ing public opinion
o nuclea ene gy on Twi e . In: P oceedings o he in e na ional
join con e ence on neu al ne wo ks. ins i u e o elec ical and
elec onics enginee s Inc.
Kim J, B ossa d D, Scheu ele DA, Xenos M (2016) “Sha ed” in o -
ma ion in he age o big da a: explo ing sen imen exp ession
ela ed o nuclea ene gy on wi e . Jou nal Mass Commun Q
93:430–445. h ps:// doi. o g/ 10. 1177/ 10776 99016 640715
Kim H (2014) 3.11 Eas Japan g ea ea hquake and wi e : ocusing
on he ela ionship be ween SNS and an i-nuclea mo emen s. 日
本硏究 273–295
LeFiga o (2021) Emmanuel Mac on annonce la cons uc ion de nou-
eaux éac eu s nucléai es de ype EPR
Li N, Akin H, Yi-Fan LS e al (2016) Twee ing disas e : An analysis
o online discou se abou nuclea powe in he wake o he Fuku-
shima Daiichi nuclea acciden . J Sci Commun 15:1–20. h ps://
doi. o g/ 10. 22323/2. 15050 202
Liu Z, Na JC (2018) Aspec -based sen imen analysis o nuclea ene gy
wee s wi h a en i e deep neu al ne wo k. In: Lec u e No es in
Compu e Science. Sp inge , pp 99–111
Mansell R (2009) The in o ma ion socie y
Mikolo T, Chen K, Co ado G, Dean J (2013) E icien es ima ion o
wo d ep esen a ions in ec o space. In: 1s In e na ional con e -
ence on lea ning ep esen a ions. pp 1–12
Miu a A, Komo i M, Ma sumu a N, Maeda K (2015) Exp ession o
nega i e emo ional esponses o he 2011 G ea Eas Japan Ea h-
quake: analysis o big da a om social media. Shin igaku Kenkyu
86:102–111. h ps:// doi. o g/ 10. 4992/ JJPSY. 86. 13076
Mo ales-i-G as J (2020a) Da os masi os y mine ía de da os sociales:
concep os y he amien as básicas. 32
Mo ales-i-G as J (2020b) Mine ía de da os de los medios sociales,
écnicas pa a el análisis de da os masi os. 33
NEI (2022) Voices o Nuclea Ene gy. h ps:// www. nei. o g/ oices- o -
nucle a - ene gy. Accessed 7 Oc 2022
Newman M, Gi an M (2004) Finding and e alua ing communi y
s uc u e in ne wo ks. Phys Re E 69:26113
NRDC (2022) Nuclea Powe 101. h ps:// www. n dc. o g/ s o i es/ nucle
a - powe - 101. Accessed 8 Oc 2022
OpenRe ine (2022) A ee, open sou ce, powe ul ool o wo king
wi h messy da a. h ps:// open e ine. o g/. Accessed 28 Ap 2021
O begozo-Te adillos J, Mo ales-i-G as J, La ondo-U e a A (2019)
Feminisms ou aged a jus ice: he online con e sa ion on “La
Manada” case. IC—Re . Cien í ica In . y Comun 2019:249–283
O begozo-Te adillos J, Mo ales-i-G as J, La ondo-U e a A (2022)
Twi e y la (de)cons ucción del mi o: Ma adona y el ac i ismo
digi al eminis a. Cuade nos.in o 52:181–203. h ps:// doi. o g/ 10.
7764/ cdi. 52. 34147
Eu opean Pa liamen (2022) Taxonomy: MEPs do no objec o inclu-
sion o gas and nuclea ac i i ies . h ps:// www. eu op a l. eu opa.
eu/ news/ en/ p ess- oom/ 20220 701IP R34365/ axon omy- meps- do-
no - objec - o- inclu sion- o - g as- an d- nucle a - ac i i ies. Accessed 23
Oc 2022
Pe iwal N (2021) Twi e Sen imen Analysis using Wo d2Vec, BiL-
STM | Kaggle. h ps:// www. kaggle. com/ code/ s oic s a ic/ wi e -
sen i men - anal y sis- using- wo d2 ec- bils m. Accessed 5 No 2022
Social Ne wo k Analysis and Mining (2023) 13:29
1 3
Page 17 o 17 29
Ran asila A, Si ola A, Kekkonen A e al (2018) # ukushima Fi e yea s
on: a mul ime hod analysis o wi e on he anni e sa y o he
nuclea disas e . In J Commun 12:928–949
Reu e s (2021) Japan’s new PM de ends p o-nuclea s ance in pa -
liamen a y debu . h ps:// www. eu e s. com/ wo ld/ asia- paci i c/
japans- new- pm- kishi da- ace- oppos i ion- ques ions- i s - ime-
2021- 10- 11/. Accessed 9 Oc 2022
Rus am F, Ash a I, Mehmood A e al (2019) Twee s classi ica ion on
he base o sen imen s o US ai line companies. En opy 21:1078.
h ps:// doi. o g/ 10. 3390/ E2111 1078
Salloum SA, Al-Em an M, Monem AA, Shaalan K (2017) A su ey o
ex mining in social media: Facebook and Twi e pe spec i es.
Ad Sci Technol Eng Sys J 2:127–133
Sano Y, To ii HA, Onoue Y, Uno K (2021) Simula ion o in o ma ion
sp eading on wi e conce ning adia ion a e he Fukushima
nuclea powe plan acciden . F on Phys 9:1–9. h ps:// doi. o g/
10. 3389/ phy. 2021. 640733
Schneide M, F ogga A (2012) 2011–2012 Wo ld nuclea indus y
s a us epo . Bull a Sci 68:8–22. h ps:// doi. o g/ 10. 1177/ 00963
40212 459126
Sen imen 140 (2022) Sen imen 140. h p:// help. sen i men 1 40. com/
home. Accessed 4 Oc 2022
Seung-Hoi K, Yu-I H, Meeyoung C, e al (2016) Public Discou se on
En i onmen al Pollu ion and Heal h in Ko ea: wee s ollowing
he Fukushima nuclea acciden . In: Social web o en i onmen al
and ecological moni o ing. pp 187–190
Sosa PM (2017) Twi e Sen imen Analysis using combined LSTM-
CNN Models. Ep in A xi 1–9
Su LYF, Caccia o e MA, Liang X e al (2016) Analyzing public sen-
imen s online: combining human- and compu e -based con en
analysis. In o ma ion Commun Soc 20:406–427. h ps:// doi. o g/
10. 1080/ 13691 18X. 2016. 11821 97
Tsuboku a M, Onoue Y, To ii HA e al (2018) Twi e use in scien i ic
communica ion e ealed by isualiza ion o in o ma ion sp ead-
ing by in luence s wi hin hal a yea a e he Fukushima Daiichi
nuclea powe plan acciden . PLoS ONE 13:1–14. h ps:// doi. o g/
10. 1371/ jou n al. pone. 02035 94
Twa c (2022) wa c. h ps:// wa c- p oje c . ead hedocs. io/ en/ la es /.
Accessed 20 Feb 2022
Twi e (2022) Twi e API o Academic Resea ch. h ps:// de el ope .
wi e . com/ en/ p odu c s/ wi e - api/ acade mic- esea ch. Accessed
20 Feb 2022
Ume M, Ash a I, Mehmood A e al (2021) Sen imen analysis o
wee s using a uni ied con olu ional neu al ne wo k-long sho -
e m memo y ne wo k model. Compu In ell 37:409–434. h ps://
doi. o g/ 10. 1111/ COIN. 12415
USDS-USEOP (2021) The Long-Te m S a egy o he Uni ed S a es,
Pa hways o Ne -Ze o G eenhouse Gas Emissions by 2050.
Washing on
Veil SR, Buehne T, Palencha MJ (2011) A wo k-in-p ocess li e a u e
e iew: inco po a ing social media in isk and c isis communica-
ion. J Con ingencies C is Manag 19:110–122. h ps:// doi. o g/ 10.
1111/J. 1468- 5973. 2011. 00639.X
Yagaha a A, Hanai K, Hasegawa S, Ogasawa a K (2018) Rela ion-
ships among wee s ela ed o adia ion: Visualiza ion using co-
occu ing ne wo ks. JMIR Public Heal Su eill 4:1–14. h ps://
doi. o g/ 10. 2196/ publi cheal h. 7598
Za abei ia-Bilbao E, Mo ales-I-G as J, Rio-Bel e RM, Ga echana-
Anacabe G (2022a) G een ene gy: Iden i ying de elopmen ends
in socie y using Twi e da a mining o make s a egic decisions.
P o La In 31:1–16. h ps:// doi. o g/ 10. 3145/ epi. 2022. ene. 14
Za abei ia-Bilbao E, Rio-Bel e R-M, Al a ez-Meaza I, Ma ínez de
Aleg ía-Mancisido I (2022b) Wo ld en i onmen day: unde -
s anding en i onmen al p og ams impac on socie y using wi e
da a mining. Soc Indic Res 164:263–284. h ps:// doi. o g/ 10. 1007/
s11205- 022- 02957-y
Publishe 's No e Sp inge Na u e emains neu al wi h ega d o
ju isdic ional claims in published maps and ins i u ional a ilia ions.
