Opinion and Sentiment Analysis of Twitter Users during the 2021

Ecuador Presidential Election

Jorge Parraga-Alava

1 a

, Jorge Rodas-Silva

2 b

, Iv

an Quimi

and Roberth Alcivar-Cevallos

1 c

Facultad de Ciencias Inform

aticas, Universidad T

ecnica de Manab

ı, Avenida Jos

e Mar

ıa Urbina, Portoviejo, Ecuador

Facultad de Ciencias e Ingenier

ıa, Universidad Estatal de Milagro, Cdla. Universitaria Km 1 1/2 v

ıa Km 26, Milagro,

Ecuador

Keywords:

Ecuador, Presidential Election, Opinion Analysis, Sentiment Analysis.

Abstract:

Social media data have been used for opinion and sentiment analysis and seem to have the potential to reﬂect

the political picture of many territories. This paper analyzes the opinions and sentiments of users about the

organization and candidates of the 2021 Ecuadorian presidential election to determine whether these can be

considered as a relevant factor to predict election outcomes in this country. We used a social media analytics

methodology with four phases: ﬁrst two correspond to data acquisition and pre-processing, where Twitter

search API was used for fetching election-related tweets that were taken and converted into a structured for-

mat; in the third phase, an opinion analysis was performed to offer statistics about the number of tweets and

users, hashtags, mentions and, word clouds. In the fourth phase, we veriﬁed the emotional attitude of the users

regarding the presidential candidates by using sentiment analysis. The results showed that most of the users’

opinions reﬂected positive sentiment about presidential candidate Arauz in the ﬁrst round. On the other hand,

in the second round, presidential candidate Lasso, concerning the ﬁrst round, captured a more signiﬁcant pos-

itive response from Twitter users, who achieved a closed result over candidate Arauz. Finally, it is concluded

that there is a correspondence between positive sentiments expressed in the tweets and the total votes obtained

by candidates.

1 INTRODUCTION

Ecuador is a country located in South America with a

population of approximately 17 million people. Pres-

idential elections in Ecuador are held every four years

since the country returned to a democracy in 1979.

Elections are mandatory for all Ecuadorian citizens

older than eighteen years old. It consists of two

rounds that are held in dates selected by Consejo Na-

cional Electoral, CNE (the state institution responsi-

ble for holding elections in the country). If any can-

didate is able to obtain more than 40 % of votes (af-

ter the count to remove invalid ballots) and if he or

she has at least 10 % over the second place is de-

clared President in the ﬁrst round. Otherwise, the two

with highest ballots go for a second round of popular

elections, where the one who gets more than 50 % of

the ballots is declared President of the nation [Rofr

ıo

et al., 2019].

https://orcid.org/0000-0001-8558-9122

https://orcid.org/0000-0001-6526-7740

https://orcid.org/0000-0001-6282-8493

Despite of the coronavirus pandemic, the 2021

Ecuador Presidential Election took place in February

7th, and the second round in April 11. During the

ﬁrst round, legislative elections were held in which

representatives to the Andean Parliament and Assem-

blymen were elected for the same period. In the ﬁrst

round, 16 binomial candidates were registered. Af-

ter the ﬁrst round, no candidate obtained the required

votes to win the elections. Therefore a second round

of elections was held in April 11, where Guillermo

Lasso was elected President of Ecuador and the run-

ner up was Andr

es Arauz, with 52.36 % and 47.64 %

of the votes, respectively

Social media have become an essential part of the

routine of the people, since they allow users to express

opinions as well as their happiness, anger, sadness, or

any other emotion easily. Such presence in people’s

lives is so great that they have been used to inﬂuence

elections in at least 18 countries, according to a report

by the democracy advocacy group Freedom House

shorturl.at/hBGT8

shorturl.at/ijmuD

Parraga-Alava, J., Rodas-Silva, J., Quimi, I. and Alcivar-Cevallos, R.

Opinion and Sentiment Analysis of Twitter Users during the 2021 Ecuador Presidential Election.

DOI: 10.5220/0010651000003058

In Proceedings of the 17th International Conference on Web Information Systems and Technologies (WEBIST 2021), pages 257-266

ISBN: 978-989-758-536-4; ISSN: 2184-3252

257

In Ecuador, eight out of ten people use social me-

dia on a daily basis. Of these, 1 million are users of

the social media called Twitter. Although, the num-

ber of Twitter users compared to the total population

of Ecuador is very small (about 6 % of the popula-

tion) [Del-Alc

azar, 2021], this social media has the

presence of very active users and mainly politically

related actors such as the politicians, public ofﬁcials,

candidates as well as news media. These actors en-

gage on the social platform as part of their political

campaigns or utilize it as a means of political deliber-

ation, advocacy and platform to exercise freedom of

speech [Parmelee, 2014].

Twitter data have been used by several studies re-

lated to political topics or general elections process in

many countries. These can be categorized as predic-

tion based and sentiment analysis based.

Works of [Singh et al., 2020, Sharma and Moh,

2016, Liu et al., 2021, Gaurav et al., 2013, Kristiyanti

et al., 2019, Wang and Gan, 2019] conducted stud-

ies to predict the outcome of elections process in

countries such as India, U.S., Venezuela, Paraguay,

Ecuador, Venezuela, Indonesia and France. These

studies use methods such as Support Vector Machine

(SVM), Naive Bayes, among others, with which they

reach prediction rates of between 74 %-90 % to es-

tablish seats in assembly, constituencies as well as

presidential winner. Despite these good results, the

limited number of tweets used in the experiments and

the particularity of each presidential election makes

that none of the work be yet able to provide a generic

method to predict the outcome of any election around

the world based upon the Twitter data.

The works of [Barnaghi et al., 2016, Gustisa

Wisnu et al., 2020, Sharma and Ghose, 2020, Jhawar

et al., 2020, Agarwal. and Bansal., 2020] performed

sentiment analysis to identify political preferences

over social media platforms. In this sense, all studies

focused on to obtain the opinion polarity of the folks

concerning general elections in diverse countries and

territories. They compare the sentiments of the users

for each of the candidates in analysis, and conclude

that, the opinion of the users is positive for the ma-

jority of candidates in comparison and furthermore,

they demonstrated that popularity in Twitter seems

to match with the election results. Also, the works

of [Troussas et al., 2016,Krouska et al., 2017,Parraga-

Alava et al., 2019] compared and evaluated classiﬁ-

cation algorithms for the sentiment analysis problem

using data from social networks. In both cases it is

evidenced the usefulness of the machine learning al-

gorithms for sentiment analysis services.

As observed in the previously mentioned works,

the topic of political opinions in Twitter is relevant

and has received substantial attention, especially for

sentiment analysis. In this sense, researchers focus

on ﬁnding opinions and recognizing the sentiments

expressed towards the general elections, politicians,

or public ﬁgures. In this paper, we analyze the opin-

ions and users sentiments from Twitter data focused

on the 2021 Ecuador Presidential Election. Our goal

is to have useful insights of the main opinions of users

about the organization of the 2021 Ecuadorian elec-

tion as well as to identify sentiments expressed by

users towards candidates and, to determine whether

these can be considered as a relevant factor to predict

election outcomes in Ecuador. The latter is relevant

because it can serve as a baseline towards the genera-

tion of a generic method to predict the results of any

election based upon the Twitter data. In this sense,

our research questions (RQs) are the following:

RQ 1: What are the main opinions expressed by

Twitter users around 2021 Ecuadorian General Elec-

tion?

RQ 2: What is the positive sentiment expressed by

Twitter users about the presidential candidates in the

Ecuadorian general elections of 2021?

RQ 3: Is there a correlation between the sentiment

expressed by the users in the tweets and the vote per-

centage obtained by the candidates?

The paper is organized as follows: starting with

an introduction about elections and Twitter as well as

related works. Section 2 offers a description of data,

techniques and software used by our analysis. The

section 3 presents the experimental results and dis-

cussion. The conclusions are given in the last section.

2 METHODS

The aim of this paper is to have useful insights of the

main opinions of users about the organization of the

2021 Ecuador Presidential Election as well as to iden-

tify sentiments expressed by user towards the candi-

dates. To accomplish this task, we follow the method-

ology as given in Figure 1.

2.1 Data Collection

We used posts on Twitter collected during February 7,

2021 (ﬁrst round) and April 11, 2021 (second round).

This period encompassed the 2021 Ecuador Presiden-

tial Election. The data were collected using the Twit-

ter API search along with rtweet R package [Kearney,

2019] and amounted to 288k tweets approximately.

The tweets acquired correspond to posts that included

the ofﬁcial hashtag promoted by the Consejo Na-

cional Electoral (CNE), that is, #Elecciones2021Ec

WEBIST 2021 - 17th International Conference on Web Information Systems and Technologies

258

01 02 03 04

DATA COLLECTION PREPROCESSING OPINION ANALYSIS SENTIMENT ANALYSIS

Twitter search API

General:

#Elecciones2021Ec #SegundaVueltaEc

Candidates:

Keywords by candidate (Table 1)

Tokenization

Stop Words

Twitter symbols

Tweets Statistics

# Analysis

@ Analysis

Word Clouds

Geolocation Analysis

# of positive and negative

tweets.

# of positive and negative

Twitter impressions.

Sentiment analysis report.

Figure 1: Pipeline of out opinion and sentiment analysis of 2021 Ecuador Presidential Election.

and #SegundaVueltaEc, for ﬁrst and second round, re-

spectively. For the presidential election candidates,

we select either of the four presidential candidates

with greater intention of the vote according to polls

Table 1 shows the keywords and ofﬁcial Twitter

proﬁles used to ﬁlter the candidates-related tweets.

2.2 Preprocessing

We preprocess the data to convert text from human

language into computer-readable format for analitycs

use. We performed the following steps for all the

tweets: tokenization, removing stop-words and Twit-

ter symbols. In tokenization process, a string is split-

ting up into a list of tokens and constructing a bag-of-

words. Thereby, each token was used and if it was any

stop word or punctuation symbol it was removed. As

in Twitter there are some symbols which may be used

in tweets and they can introduce meaningless noise,

the URLs, “RTs”, character representing strip white

space and emojis are removed. We keep emojis out

from the analysis because the number of tweets with

emojis in our dataset was small (only 5 % and mostly

the ballot box with ballot emoji). We used the tidy-

verse [Wickham et al., 2019] and tm [Feinerer et al.,

2008] R packages for this stage. In the end, this pre-

processing converts the tweets into a dataset where

each row contains status id, user id, created at, screen

name, text, place name and country related to a Twit-

ter post. These features will be used to carry out the

analysis of opinions and sentiments.

2.3 Opinions Analysis

To extract opinions from a posts in Twitter, we ap-

plied two social media analytics techniques, namely

shorturl.at/lvDIN

descriptive analysis and geospatial analysis. The

descriptive analysis gives descriptive statistics about

number of tweets, number of tweet users, hashtags,

mentions and word clouds [Singh et al., 2020]. The

geospatial analysis deals with the study the topics

spread by the user throughout geographic areas [Ban-

thia et al., 2020]. We used the tidyverse [Wickham

et al., 2019] and wordcloud

R packages to carry out

both descriptive and geospatial analysis, respectively.

2.4 Sentiment Analysis

We performed an analysis process based on the sen-

timents got from posts on Twitter related to the pres-

idential candidates. We quantiﬁed these sentiments

using polarity. Polarity analysis is used to determine

the emotional attitude of a text writer with respect to

the topic under discussion [Li and Wu, 2010]. With

it, the text of a tweet can be classiﬁed as negative,

positive or neutral. It assigns scores range from -1 to

1, where -1 represents extremely negative sentiment

while 1 represents extremely positive sentiment, re-

spectively. A polarity score of 0 suggests a neutral

sentiment [Yaqub et al., 2020].

We quantify the sentiments of Twitter users us-

ing their polarity towards each of the candidates. In

ﬁrst round, we focus solely on the presidential can-

didates and we classiﬁed tweets as Arauz-related,

Lasso-related, P

erez-related or Hervas-related. We

leave out the tweets which mention more than two

candidates because of the potential ambiguity as to

which candidate the sentiment of the tweet is about.

In second round, we only consider the two ﬁnalist

candidates.

We used the Phyton library Tweepy

for perform a

shorturl.at/mIRX4

shorturl.at/kpEPZ

Opinion and Sentiment Analysis of Twitter Users during the 2021 Ecuador Presidential Election

259

Table 1: Keywords for ﬁltering the tweets by presidential candidates.

Candidate Twitter proﬁle Keyword

Andr

es Arauz @ecuarauz

arauz, #ArauzPresidente2021

#ArauzPresidente, #EcuadorConArauz

Guillermo Lasso @LassoGuillermo

lasso, #LassoPresidente2021

#LassoPresidente, #EcuadorConLasso

Yaku P

erez @yakuperezg

yaku, #EcuadorConYaku

#YakuPresidente, #YakuPresidente2021

Xavier Hervas @xhervas

hervas, #EcuadorConHervas

#HervasPresidente, #HervasPresidente2021

sentiment analysis and plot the result. The text analy-

sis to identify, extract and display subjective informa-

tion about the presidential candidates (good, bad, ex-

cellent, lousy), was performed through the TextBlob

and Matplotlib

libraries, also included in Phyton.

With the application of these libraries on the infor-

mation extracted from Twitter, an approximation of

the emotional evaluation that voters had of the presi-

dential candidates in the last elections held in Ecuador

was achieved.

3 RESULTS AND DISCUSSION

3.1 Opinions Analysis

To answer the RQ 1, we performed a descriptive

analysis and a geospatial analysis. In the ﬁrst one,

we included tweet statistics about number of tweets

and number of tweet users that they posted during

the elections, hashtags and mentions analysis as well

as word clouds of the opinions of the user. In the

last one, we identiﬁed from which province of the

Ecuadorian territory the tweets with mentions of each

candidate and process in general were posted.

3.1.1 Tweet Statistics

The detailed results of tweet statistics are shown in

Table 2.

Here, among 84955 tweets, 68150 (80.21 %)

tweets were in ﬁrst round and 16805 (19.79 %) in

second round, related to main opinions of the election

(#Elecciones2021Ec and #SegundaVueltaEc). These

post were generated by a total of 34983 unique users,

distributed in 28039 (80.15 %) and 6944 (19.85 %),

for ﬁrst and second round, respectively.

When the tweets with mentions of candidates are

analyzed, it is observed that a total of 203056 were

obtained from Twitter. They are distributed as fol-

shorturl.at/pELY2

shorturl.at/gqrCJ

lows: 88589 (43.62 %), 99754 (49.12 %), 7562 (3.73

%), 7151 (3.53 %) for Andr

es Arauz, Guillermo

Lasso, Yaku P

erez and Xavier Hervas, respectively.

These post were generated by a total of 95316 unique

users, distributed in 43843 (45.99 %), 45212 (47.43

%), 2541 (2.66 %) and 3720 (3.92 %), for Andr

Arauz, Guillermo Lasso, Yaku P

erez and Xavier Her-

vas, respectively.

Our tweets statistics analysis showed that more

tweets were generated with mentions of the presiden-

tial candidates than the process in general. In this

sense, to refer to the main opinions of the election pro-

cess, each user posted an average of 3 tweets. While

each user who mentions to Arauz, Lasso, P

erez and

Hervas posted an average of 2, 3, 3, and 2 tweets,

respectively. Regarding the ﬁnalist candidates, it is

observed that, Andr

es Arauz had a decrease of almost

12 % of mentions in the second round. In contrast,

Guillermo lasso was who had a signiﬁcant increase in

mentions in tweets, going from 28950 tweets in ﬁrst

round to 70804 tweets in the second round, which rep-

resents an increase of 59.11 %.

3.1.2 Hashtag Analysis

A total of 788 hashtags were identiﬁed in 84955

tweets in approximately 80 % (68150) of tweets for

ﬁrst round and 20 % (16805) for second round related

to opinions of user about general process. The top 15

hashtags that had maximum occurrences in tweets in

two rounds are shown in Figure 2.

The Figure 2A highlights the hashtags #VotaSe-

guro, #YoMeCuido, #JuntosContraElCovid19 and

#NoBajemosLaGuardia directly related to user tips,

suggestions and complaints on aspects related to pol-

icy on conducting elections in the context of the

Covid-19 pandemic. In Figure 2B, however, these

topics decreased and only the hashtag #VotaSeguro

remained, and other hashtags related to the election

process and presidential candidates (#LassoYaNos-

Goberno, #LassoPresidente2021) stand out.

Our hashtag analysis evidenced that #EcuadorDe-

cide2021 and #LassoPresidente2021 were the most

WEBIST 2021 - 17th International Conference on Web Information Systems and Technologies

260

Table 2: Tweet statistics of data used in the analysis.

Statistics General

Presidential candidate

Arauz Lasso P

erez Hervas Total

Total Tweets 100.00 % 43.62 % 49.12 % 3.73 % 3.53 % 100.00 %

- Total Tweets (ﬁrst round) 80.21 % 53.23 % 31.01 % 8.10 % 7.66 % 100.00 %

- Total Tweets (second round) 19.79 % 35.45 % 64.55 % - - 100.00 %

Total Unique Users 100.00 % 45.99 % 47.43 % 2.66 % 3.92 % 100.00 %

- Total Unique Users (ﬁrst round) 80.15 % 54.04 % 30.47 % 6.28 % 9.20 % 100.00 %

- Total Unique Users (second round) 19.85 % 40.07 % 59.93 % - - 100.00 %

2392

2194

1986

1469

1462

1326

1089

1037

968

928

887

835

786

673

650

First Round

0 500 1000 1500 2000 2500

#ECU911

#LassoPresidente2021

#DENUNCIA

#JuntosContraElCovid19

#sorrynotsorry

#NoBajemosLaGuardia

#EcuadorDecide

#Quito

#Manabi

#YoMeCuido

#ATENCION

#LassoYaNosGoberno

#Guayaquil

#VotaSeguro

#EcuadorDecide2021

1683

430

378

273

253

249

227

225

208

191

187

185

184

173

164

Second Round

0 500 1000 1500

#LassoPresidente2021

#EcuadorDecide

#Quito

#ATENCION

#Guayaquil

#VotaSeguro

#EcuadorDecide2021

#LassoPresidente

#CNE

#EleccionesBicentenario

#GuillermoLasso

#ZapatosRojos

#Arauz

#Lasso

#EleccionesGenerales2021

Figure 2: Top 15 hashtags with maximum occurrence related to general perception of 2021 Ecuador Presidential Election.

popular hashtags with a total of 2072 and 1377 tweets,

respectively, when we focus on the general perception

of users about the the elections. The presence in the

ﬁrst place of #LassoPresidente2021 in the global data

can be explained by the election of Guillermo Lasso

as president of Ecuador for the period 2021-2025

3.1.3 Mention Analysis

A total of 35066 mentions were present in the second

round, out of which 13145 were for Andr

es Arauz y

21921 for Guillermo Lasso. The top 10 mentions that

had maximum occurrences in the second round are

shown in Figure 3. In the case of the @ecuarauz,

the mentions (Figure 3A) are led by @UdlaChann-

shorturl.at/pyGUW

elEc who made a large number of publications related

to news about the electoral process, especially about

the counting of votes around the country in its dif-

ferent provinces. The user @uscocovich1972, who

occupies the second place, and these tweets offer in-

formation about the candidate Andres Arauz, debat-

ing his background and political party. The tweets are

informational only and do not appear to bias towards

any of the ﬁnalist candidates. Users who mention the

candidate Andr

es Arauz, generally also mention the

@MashiRafael, which is the account of the former

president Rafael Correa.

For the candidate Guillermo Lasso (Figure 3B),

the mentions are led by the account @codes

r who

presents strong support for the candidate, although

most of his posts are repetitions of the phrase “Sin

Opinion and Sentiment Analysis of Twitter Users during the 2021 Ecuador Presidential Election

261

112

@LassoGuillermo

@ecuarauz

@FEscrutinio

@Datoworld

@EduardBernstei1

@01Rome

@CanalRTU

@SebastianChawpi

@briannjavierc01

@higuerahernan

@uscocovich1972

@UdlaChannelEc

@anibalvv1982

@antibanqueroshp

@gardfieldking

@titireyesm

@JarabaWilson

@magusreyes

@mbtello

@joselo681513

@roger_gonzalezg

@codes_r

0 20 40 60

0 30 60 90

Figure 3: Top 10 mentions with maximum occurrence related to candidates of the second round of 2021 Ecuador Presidential

Election.

conﬁarnos vamos con fuerza hasta el triunfo debemos

cuidar los votos la victoria est

a cercana Lasso Pres-

idente”. The second proﬁle that mentions @Guiller-

moLasso the most is @roger gonzalezg with tweets

related to the hope of a prosperous country and want-

ing to change. Users who mention the candidate

Guillermo Lasso, generally also mention the @Cre-

oEcuador and @La6Ecuador, which are the accounts

of the candidate’s allied political parties.

Our mentions analysis evidenced that Andr

Arauz has great support from users, although most of

them are related to the party behind this candidate.

The mentions of Guilermo Lasso were related to the

hope of change to have a government that helps solve

corruption, crime and the economic crisis caused by

the pandemic. It was also evidenced that Arauz’s

detractors directly related him to former Ecuadorian

president Rafael Correa and to being his continuity,

while Guillermo Lasso’s detractors pointed to him as

responsible for the economic crisis in Ecuador in the

decade of the 90 when he was Minister of Economy.

3.1.4 Word Clouds

In Figure 4 and Figure 5 we show words with the

highest frequencies in tweets where users express

their opinions about general electoral process during

the ﬁrst and second round, respectively. Here, larger

the word more is its frequency in the given text.

Figure 4 shows words with minimum frequency

of 1214. In this sense, 100 words were selected with

word “ecuador” having maximum frequency of 16523

followed by word “arauz” with frequency of 10823.

Figure 5 shows words with minimum frequency

of 24. In this sense, 100 words were selected with

word “electoral” having maximum frequency of 304

followed by word “ecuador” with frequency of 275.

Our word clouds evidenced that the most frequent

words of the ﬁrst electoral round referred to the vot-

arauz

voto

electoral

resultados

lasso

votar

andres

electorales

cne

votacion

derecho

recintos

pais

vuelta

eleccionesecuador

proceso

guillermo

candidato

yaku

jornada

ecuadordecide

bioseguridad

guayaquil

ciudadania

gente

segunda

correa

atencion

denuncia

manabi

ecuatorianos

votaseguro

perez

medidas

democracia

hacer

estan

ahora

ejercer

papeletas

votos

nacional

junto

oficiales

hervas

filas

recinto

correismo

ciudadanos

lugar

presidencial

centros

quito

exit

personas

presidente

pueblo

colegio

poll

voluntad

dia

urnas

actas

fila

conteo

mientras

anos

seguridad

libertad

transparencia

presidenciales

popular

sectores

lassoyanosgoberno

cuenta

elegir

estrategia

populares

sistema

social

futuro

mundo

yomecuido

juntas

participacion

delegados

atentos

puntos

domingo

viendo

mascarilla

patria

derecha

primer

febrero

rapido

mediante

datos

sufragar

manana

mucha

preferencia

centro

preliminares

elegido

esperanza

cumpli

expresar

seguido

diferencia

democrata

muestra

defiende

ecuatoriano

debida

mesa

van

honre

votantes

fraude

parte

momento

via

autoridades

largas

gobierno

puede

guayas

evidenciar

aqui

receptoras

representantes

primera

venezuela

vote

presidenta

bien

sucediendo

gano

denunciamos

retardar

mesas

aglomeraciones

ciudadana

papeleta

amigos

segun

joven

electores

ecuador

Figure 4: Word cloud of frequent words in posts about gen-

eral perception during ﬁrst round of the 2021 Ecuador Pres-

idential Election.

jor

lasso

arauz

voto

ahora

andres

guillermo

actas

proceso

electorales

resultados

recintos

cne

eleccionesgenerales

escrutinio

provincial

votos

desarrollo

junta

votacion

derecho

nacional

provincia

seguridad

procesadas

normalidad

presidente

inauguracion

hrs

bioseguridad

quito

pais

lassopresidente

ciudadania

oficiales

eleccionesecuador

autoridades

vuelta

sesion

sufragio

juntas

medidas

abril

receptoras

detalles

permanente

presidenta

democracia

recinto

segunda

traves

votar

ecuatorianos

presencia

urnas

centro

electores

director

ecuasondeos

cneinforma

desarrolla

datos

mesa

acto

candidato

ejercer

guayaquil

ciudadanos

delegados

ecuadordecide

educativa

personas

unidad

vicepresidente

estan

junto

atencion

mision

participacion

total

corte

dia

habilitados

republica

sufragar

colegio

instituciones

transmision

canton

delegacion

diferentes

domingo

medios

procesamiento

politicas

vivo

preliminares

via

electoral

ecuador

nada

Figure 5: Word cloud of frequent words in posts about gen-

eral perception during second round of the 2021 Ecuador

Presidential Election.

ing, the objective of the presidency, requests for le-

gal suffrage, biosafety measures and ofﬁcial results,

as well as the mentions of each candidate. Mean-

while for second electoral round The word cloud of

the second round similar to what happens in the ﬁrst

electoral round with respect to electoral journey but

topics such as transparent process, scrutiny and ofﬁ-

WEBIST 2021 - 17th International Conference on Web Information Systems and Technologies

262

100

@GuillermoLasso

First

Round

Second

Round

User posts (%)

7.3

0.3

1.4

2.8

3.1

34.6

0.3

2.4

0.3

3.5

0.3

36.7

0.3

0.7

3.1

6.5

0.2

0.6

0.2

0.5

2.7

1.2

42.9

2.1

1.8

5.8

0.3

0.1

0.3

0.2

1.1

0.1

1.3

0.1

0.3

3.3

0.7

36.7

1.7

0.3

4.7

0.3

1.7

0.7

0.3

1.3

100

6.6

0.3

1.4

0.5

1.4

0.3

2.4

1.2

37.5

1.8

1.4

1.7

7.4

0.2

0.1

32.2

0.6

0.5

0.3

1.9

100

@ecuarauz

First

Round

Second

Round

User posts (%)

Figure 6: Percentages of tweets for ﬁnalist candidates in provinces of Ecuador in the 2021 Ecuador Presidential Election.

cial results as well as the mention of the ﬁnalist can-

didates are added. Note that the frequency of words

in the second round decreased signiﬁcantly and this is

due to the fact that in the second round users focused

more on the candidates than on the electoral process

itself.

3.1.5 Geospatial Analysis

Figure 6 shows the spread by the user posts through-

out geographic area of Ecuador. the total of tweets

with geolocation, 48.5 % refer to Andr

es Arauz and

51.5 % to Guillermo Lasso. Note that only the two

ﬁnalist candidates were considered because the num-

ber of tweets with geolocation for the other candidates

was extremely low, i.e., less than 0.05 % of total of

their tweets.

In the case of the ﬁrst round, the posts with men-

tions or references to Andr

es Arauz (@ecuarauz)

came mainly from the provinces of Guayas (36.7 %)

and Pichincha (39 %), adding together more than 75

% of the posts, and there were no postings from 5

provinces of Ecuador, mostly from the Andean and

Amazon region. For Guillermo Lasso (@Guillermo-

Lasso), there is a very similar posting behavior by

provinces, that is, the majority of tweets come from

Guayas (34.6 %) and Pichincha (36.7 %)

In the second round, for Andr

es Arauz (@ecua-

rauz) the majority of mentions in Guayas (37.5 %)

and Pichincha (32.2 %) continue to be maintained, but

an increase is noted in the province of Manab

ı, where

it went from 4.7 % of tweets generated in the ﬁrst

round to 7.4 % in the second. It is also observed that

the number of provinces without posts was reduced

to only two. Regarding Guillermo Lasso (@Guiller-

moLasso), no signiﬁcant changes are observed with

respect to the ﬁrst round in terms of the provinces

from which more tweets were generated. However,

it is evident that on this occasion in all the provinces

mentions of the candidate were generated, probably

due to his triumph as president of Ecuador.

Our geospatial analysis evidenced that the most of

the tweets with geolocation and mentions of the two

candidates came from the provinces of Guayas and

Pichincha. In the case of Andr

es Arauz (@ecuarauz),

the provinces of Manab

ı, Azuay and El Oro also stand

out, to a lesser extent. In the case of Guillermo Lasso

(@GuillermoLasso), on the other hand, they are the

provinces of Manab

ı, Azuay and Loja.

3.2 Sentiment Analysis

To answer the RQ 2, we performed the following data

analysis: With the data collected from Twitter, about

the presidential elections, we deﬁned the tags derived

from hashtags to train the sentiment classiﬁers of the

tool used in this work and to know the perception of

users about candidates. This process was performed

in the ﬁrst and second round. Because of our dataset

does not have neutral data, the tags used to know the

sentiment about the candidates were frequent: “Good

president”, “Bad president”, “good”, “bad”, “lousy”,

“worst government”, “good government”.

With tags deﬁned, we ran the algorithm included

in the tool described in 2.4, which performs a classiﬁ-

cation of the data by dividing them into groups of pos-

itive and negative. For this work we considered the

positive sentiment results because the objective was

to know the favorable impact each candidate had on

users. Figure 7 shows the positive sentiment results

obtained for each candidate in the ﬁrst round.

Opinion and Sentiment Analysis of Twitter Users during the 2021 Ecuador Presidential Election

263

Results of Arauz's sentiment on Twitter

Results of Lasso's sentiment on Twitter

Results of Hervas's sentiment on Twitter

Results of Perez's sentiment on Twitter

Number of tweets Number of tweets

Sentiment

Average sentiment :

Round

32,56 %

16,28 %

25,58 %

Figure 7: Sentiment results in the ﬁrst round of the presidential election.

Results of Arauz's sentiment on Twitter

Results of Lasso's sentiment on Twitter

Average sentiment :

Number of tweets Number of tweets

Sentiment

Round

51,85 %

48,12 %

Figure 8: Sentiment results in the second round of the presidential election.

As shown in ﬁgure 7, presidential candidates

Arauz, Lasso, Hervas and P

erez, were the ones who

generated the most information trafﬁc on Twitter, pro-

voking different points of opinion among users. For

the case of Arauz, we can observe that he obtained

an average positive value of 32,56 %, Lasso 16,28 %,

while Perez and Hervas an average of 25,58 %. Arauz

was candidate who caused more sympathy among

WEBIST 2021 - 17th International Conference on Web Information Systems and Technologies

264

Table 3: Votes and Positives Sentiments obtained by candidates of the 2021 Ecuador Presidential Election.

Candidate

First round Second round

Positive Sentiment Total Votes Positive Sentiment Total Votes

Andr

es Arauz 32.56 % 32.72 % 51.85 % 47.64 %

Guillermo Lasso 16.28 % 19.74 % 48.15 % 52.36 %

Yaku P

erez 25.58 % 19.39 % - -

Xavier Hervas 25.58 % 15.68 % - -

users which leads us to say that the majority agreed

with the electoral tweets that generated trends during

the ﬁrst round. These results were contrasted with

the ofﬁcial electoral results obtained in the ﬁrst round

where Arauz, Perez, Lasso and Hervas were those

who occupied the ﬁrst positions in the elections re-

spectively.

In the second round, the same process described

in the ﬁrst round was carried out, results of positive

sentiments are shown in Figure 8. The ﬁgure shows

the perception of candidates Arauz and Lasso, if we

compare them with the results obtained by both can-

didates in the ﬁrst round, in this case the results were

similar. Arauz obtained an average positive value of

51,85 % while Lasso obtained 48,12 %. Results ob-

tained evidence the positive sentiment of the tweets

that users posted on Twitter during the electoral con-

test in the second round.

3.3 Correlation between Sentiments and

Votes

To answer the RQ 3, we show in Table 3, the positive

sentiments expressed by users towards each candidate

with the votes ﬁnally obtained by them in the ofﬁ-

cial elections. Here, the total votes are obtained from

the CNE website for ﬁrst

and second round

. From

these data, we computed the Pearson correlation co-

efﬁcient (ρ) to measure the statistical relationship be-

tween two variables. Here, we considered as variables

X = positive sentiments and Y = total votes. We com-

puted ρ and we obtained a value of 0.63 for the ﬁrst

round. This indicates that there tends to be a positive

association, i.e., as positive sentiment increases, total

votes tends to increase. For the second round, it is not

possible to compute the coefﬁcient because it require

at least 3 points data to offer a meaningful capture

about the linear correlation.

Our simple correlation analysis evidenced that the

positive sentiments expressed towards the presiden-

tial candidates are similar to the votes ﬁnally obtained

during the ﬁrst round of the election. However, it is a

way simple method used for a single election process.

shorturl.at/jsIV0

shorturl.at/xAX29

Furthermore, we observe that for the second round

the percentages of positive sentiments are not as pre-

cise (as in the ﬁrst round) in relation to the total votes

obtained by the presidential candidates. Therefore,

we believe that this correlation analysis requires more

historical data from other elections and a deeper anal-

ysis of the country’s social context, to come to con-

sider Twitter sentiment analysis as a relevant factor to

predict election outcomes in Ecuador.

4 CONCLUSIONS

This paper presented an opinion and sentiment anal-

ysis of Twitter users during the 2021 Ecuador Presi-

dential Election. It includes an interesting look into

the sentiments expressed by users and their relation

with the ofﬁcial voted achieved by presidential candi-

dates.

Opinions analysis shows that users offered opin-

ions on biosecurity measures due to the pandemic and

speed in providing the ofﬁcial results quickly during

the ﬁrst round. In the second round, opinions were

more focused on mentioning positive and negative as-

pects of each ﬁnalist candidate.

Sentiment analysis shows results of user’s perception

of the candidates from tags extracted from tweets col-

lected from Twitter. Results showed that candidate

Arauz was the presidential candidate who obtained

the most positive sentiment in the ﬁrst and second

round. However, for the second round, results ob-

tained by presidential candidate Lasso evidenced a

close relationship of positive sentiment over Arauz.

The ﬁnal results of the sentiment analysis on Twitter

about Ecuador’s presidential elections in 2021 were

close to the ofﬁcial results published by the CNE.

Correlation analysis demonstrated that there is a cor-

respondence between positive sentiments expressed

in the tweets and the total votes obtained by candi-

dates during the ﬁrst round of the 2021 Ecuador Pres-

idential Election. However, it does not seem sufﬁcient

to produce a reliable result regarding sentiments anal-

ysis in Twitter to be considered a relevant factor in

predicting election outcomes in Ecuador.

Opinion and Sentiment Analysis of Twitter Users during the 2021 Ecuador Presidential Election

265

REFERENCES

Agarwal., A. and Bansal., V. (2020). Exploring sentiments

of voters through social media content: A case study

of 2017 assembly elections of three states in india. In

Proceedings of the 22nd International Conference on

Enterprise Information Systems - Volume 1: ICEIS,,

pages 596–602. INSTICC, SciTePress.

Banthia, M., Goel, S., Deepanshi, and Sinha, A. (2020).

Geo-spatial analysis of information outreach in twitter

social network. In Singh, P. K., Sood, S., Kumar, Y.,

Paprzycki, M., Pljonkin, A., and Hong, W.-C., editors,

Futuristic Trends in Networks and Computing Tech-

nologies, pages 444–455, Singapore. Springer Singa-

pore.

Barnaghi, P., Ghaffari, P., and Breslin, J. G. (2016). Opinion

Mining and Sentiment Polarity on Twitter and Corre-

lation between Events and Sentiment. Proceedings -

2016 IEEE 2nd International Conference on Big Data

Computing Service and Applications, BigDataService

2016, pages 52–57.

Del-Alc

azar, J. (2021). Ecuador Estado Digital Ene/21.

Mentinno - Formaci

on Gerencial - New Media.

Feinerer, I., Hornik, K., and Meyer, D. (2008). Text min-

ing infrastructure in r. Journal of Statistical Software,

25(5):1–54.

Gaurav, M., Kumar, A., Srivastava, A., and Miller, S.

(2013). Leveraging candidate popularity on Twitter

to predict election outcome. Proceedings of the 7th

Workshop on Social Network Mining and Analysis,

SNA-KDD 2013.

Gustisa Wisnu, G. R., Ahmadi, Muttaqi, A. R., Santoso,

A. B., Putra, P. K., and Budi, I. (2020). Sentiment

analysis and topic modelling of 2018 central java gu-

bernatorial election using twitter data. 2020 Interna-

tional Workshop on Big Data and Information Secu-

rity, IWBIS 2020, pages 35–40.

Jhawar, A., Munjal, V., Ranjan, S., and Karmakar, P.

(2020). Social Network based Sentiment and Net-

work Analysis to Predict Elections. Proceedings of

CONECCT 2020 - 6th IEEE International Conference

on Electronics, Computing and Communication Tech-

nologies, pages 0–5.

Kearney, M. W. (2019). rtweet: Collecting and analyzing

twitter data. Journal of Open Source Software, 4(42).

R package version 0.7.0.

Kristiyanti, D. A., Normah, and Umam, A. H. (2019). Pre-

diction of Indonesia presidential election results for

the 2019-2024 period using twitter sentiment analysis.

Proceedings of 2019 5th International Conference on

New Media Studies, CONMEDIA 2019, pages 36–42.

Krouska, A., Troussas, C., and Virvou, M. (2017). Compar-

ative evaluation of algorithms for sentiment analysis

over social networking services. Journal of Universal

Computer Science, 23(8):755–768.

Li, N. and Wu, D. D. (2010). Using text mining and senti-

ment analysis for online forums hotspot detection and

forecast. Decision Support Systems, 48(2):354–368.

Liu, R., Yao, X., Guo, C., and Wei, X. (2021). Can We

Forecast Presidential Election Using Twitter Data?

An Integrative Modelling Approach. Annals of GIS,

27(1):43–56.

Parmelee, J. H. (2014). The agenda-building function of

political tweets. New Media & Society, 16(3):434–

450.

Parraga-Alava, J., Caicedo, R. A., G

omez, J. M., and

Inostroza-Ponta, M. (2019). An unsupervised learn-

ing approach for automatically to categorize potential

suicide messages in social media. In 2019 38th Inter-

national Conference of the Chilean Computer Science

Society (SCCC), pages 1–8.

Rofr

ıo, D., Ruiz, A., Sosebee, E., Raza, Q., Bashir, A.,

Crandall, J., and Sandoval, R. (2019). Presidential

elections in ecuador: Bot presence in twitter. In 2019

Sixth International Conference on eDemocracy eGov-

ernment (ICEDEG), pages 218–223.

Sharma, A. and Ghose, U. (2020). Sentimental Analysis

of Twitter Data with respect to General Elections in

India. Procedia Computer Science, 173(2019):325–

334.

Sharma, P. and Moh, T. S. (2016). Prediction of Indian elec-

tion using sentiment analysis on Hindi Twitter. Pro-

ceedings - 2016 IEEE International Conference on

Big Data, Big Data 2016, pages 1966–1971.

Singh, P., Dwivedi, Y. K., Kahlon, K. S., Pathania, A., and

Sawhney, R. S. (2020). Can twitter analytics predict

election outcome? An insight from 2017 Punjab as-

sembly elections. Government Information Quarterly,

37(2):101444.

Troussas, C., Krouska, A., and Virvou, M. (2016). Evalua-

tion of ensemble-based sentiment classiﬁers for twit-

ter data. In 2016 7th International Conference on In-

formation, Intelligence, Systems Applications (IISA),

pages 1–6.

Wang, L. and Gan, J. Q. (2019). Prediction of the 2017

French Election Based on Twitter Data Analysis Us-

ing Term Weighting. 2018 10th Computer Science

and Electronic Engineering Conference, CEEC 2018

- Proceedings, pages 231–235.

Wickham, H., Averick, M., Bryan, J., Chang, W., Mc-

Gowan, L. D., Franc¸ois, R., Grolemund, G., Hayes,

A., Henry, L., Hester, J., Kuhn, M., Pedersen, T. L.,

Miller, E., Bache, S. M., M

uller, K., Ooms, J., Robin-

son, D., Seidel, D. P., Spinu, V., Takahashi, K.,

Vaughan, D., Wilke, C., Woo, K., and Yutani, H.

(2019). Welcome to the tidyverse. Journal of Open

Source Software, 4(43):1686.

Yaqub, U., Sharma, N., Pabreja, R., Chun, S. A., Atluri, V.,

and Vaidya, J. (2020). Location-based sentiment anal-

yses and visualization of twitter election data. Digit.

Gov.: Res. Pract., 1(2).

WEBIST 2021 - 17th International Conference on Web Information Systems and Technologies

266