BATMAN: A Big Data Platform for Misinformation Monitoring

Ivandro Claudino, Luciano Galic, Wellington Franco, Thiago Gadelha, Jos

e Maria Monteiro

and Javam Machado

Universidade Federal do Cear

a, Brazil

jose.monteiro@lsbd.ufc.br, javam.machado@lsbd.ufc.br

Keywords:

Big Data, Misinformation Monitoring, Platform, Social Networks.

Abstract:

The large-scale dissemination of misinformation through social media has become a critical issue, harming

social stability, democracy, and public health. The WhatsApp instant messaging application is very popular in

Brazil, with more than 165 million users. On the other hand, in just one year, the proportion of smartphones

with Telegram installed grew in Brazil from 45% to 60% in 2022. If on one hand, these platforms offer security

and privacy to its users, on other hand they are spaces with little or no moderation. Consequently,they have

been used to spread misinformation. In this context, we present BATMAN, a Big Data Platform for Misinfor-

mation Monitoring, a real-time platform for ﬁnding, gathering, analyzing, and visualizing misinformation in

social networks, in particular, in instant message applications such as WhatsApp and Telegram. To evaluate

the proposed platform, we used it to build two different messages datasets, concerning the Brazilian general

elections campaign in 2022, obtained from public chat groups on WhatsApp and Telegram, respectively.

1 INTRODUCTION

Lately, the popularity of instant messaging applica-

tions, such as WhatsApp and Telegram, has con-

tributed to the spread of misinformation. Through

these systems, misinformation can deceive thousands

of people in seconds and cause signiﬁcant harm to

individuals or society. Such platforms allow content

to be spread without editorial judgment. In this con-

text, misinformation has been used to change political

scenarios, to spread ineffective treatments, and even

to cause deaths (Martins et al., 2022; Martins et al.,

2021a; Silva and Benevenuto, 2021).

The WhatsApp instant messaging application is

very popular in Brazil, with more than 165 million

users in about 214 million people. In Brazil, more

than 95% of smatphone users use WhatsApp daily

and 48% of the population use WhatsApp to get, share

and discuss news (de S

a et al., 2021; Newman et al.,

2021). On the other hand, in just one year, the pro-

portion of smartphones with Telegram installed grew

in Brazil from 45% to 60% in 2022. The popularity

of these platforms is due to the versatility and ease of

use. They make it possible to instantly share differ-

ent media types, such as images, audios, and videos.

Besides, they provide a signiﬁcant feature: the pub-

lic chat groups. These public groups are accessible

through invitation links and, usually, they have spe-

ciﬁc topics for discussion, such as politics and edu-

cation. Both WhatsApp and Telegram, allow users

to join or even share their public groups to simulta-

neously connect to hundreds of people at once, and

quickly receive and share digital content.

In this context, monitoring the content that circu-

lates in public chat groups is a fundamental task to

understand the misinformation spreading and get in-

sights to address this problem. However, collecting

a database of messages already in circulation in chat

public groups is a challenging task. To ﬁll this gap, we

built the BATMAN, a Big Data Platform for Misinfor-

mation Monitoring, which supports ﬁnding, gather-

ing, analyzing, and visualizing misinformation in dif-

ferent social networks, in particular, in instant mes-

sage applications such as WhatsApp and Telegram.

To evaluate the proposed platform, we used it to build

two different datasets, concerning the Brazilian gen-

eral elections campaign in 2022, obtained from public

chat groups on WhatsApp and Telegram, respectively.

The remainder of this paper is organized as fol-

lows. Section 2 presents the main related work. Sec-

tion 3 describes the BATMAN platform. Section 4 de-

tails a case study performed to evaluate the proposed

platform. Conclusions and future work are presented

in Section 5.

Claudino, I., Galic, L., Franco, W., Gadelha, T., Monteiro, J. and Machado, J.

BATMAN: A Big Data Platform for Misinformation Monitoring.

DOI: 10.5220/0011995500003467

In Proceedings of the 25th International Conference on Enterprise Information Systems (ICEIS 2023) - Volume 1, pages 237-246

ISBN: 978-989-758-648-4; ISSN: 2184-4992

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

237

2 RELATED WORK

Despite the scientiﬁc community’s efforts, there is

still a need for monitoring and identifying misinfor-

mation in WhatsApp and Telegram messages, mainly

in Portuguese. The paper presented in (Garimella

and Tyson, 2018) is a seminal work in collecting and

analyzing WhatsApp’ messages. The authors built

a dataset by crawling 178 public groups, containing

45K users and 454K messages, from different coun-

tries and languages, such as India, Pakistan, Rus-

sia, Brazil, and Colombia. In the study presented in

(Machado et al., 2019), the authors collected and an-

alyzed 298,892 WhatsApp’ messages, from 130 pub-

lic groups, in the period leading up to the two rounds

of the 2018 Brazilian presidential elections. In (Re-

sende et al., 2019), the authors analyzed different as-

pects of WhatsApp messages from public political-

oriented groups. The messages were collected during

major social events in Brazil: a national truck drivers’

strike and the Brazilian presidential campaign. The

authors analyzed the types of content shared within

such groups and the network structures that emerge

from user interactions.

In (Resende et al., 2018), the authors presented a

system for gathering, analyzing, and visualize pub-

lic groups in WhatsApp. Besides describing their

methodology, the authors also provide a brief char-

acterization of the 169.154 messages shared by 6,314

users in 127 public groups to help journalists and re-

searchers understand the repercussion of events re-

lated to the 2018 Brazilian elections. In (de S

a et al.,

2021), the authors presented the Digital Lighthouse,

an entire platform for ﬁnding, gathering, analyzing,

and visualize public groups in WhatsApp. In (Cabral

et al., 2021) the author built a large-scale, labeled,

anonymized, and public dataset formed by What-

sApp messages in Portuguese (PT-BR), concerning

the Brazilian general elections campaign in 2018, col-

lected from public chat groups, using the platform

proposed by (de S

a et al., 2021). Then, the authors

conducted a series of classiﬁcation experiments us-

ing combinations of Bag-Of-Words features and clas-

sical machine learning methods, resulting in a total

of 108 experiments, in order to build a specif MID

for WhatsApp messages. Their best results achieved

a F1-score of 0.733, which served as a baseline for

other work. As a practical result of this work, the au-

thors built and deployed a Misinformation Detector,

which receives a text as input and returns as output

the probability that the text contains some misinfor-

mation

Currently, the Misinformation De-

tector can be accessed through the link

https://faroldigital.info/classiﬁer/misinformation-text

In (Martins et al., 2021a), the authors presented

a large-scale, labeled, and public data set of What-

sApp messages in Brazilian Portuguese about coro-

navirus pandemic, called COVID-19.BR, which was

collected from public chat groups, using the platform

proposed by (de S

a et al., 2021). In that work, they

conduct a series of classiﬁcation experiments using

nine different machine learning methods to build an

efﬁcient misinformation classiﬁer for WhatsApp mes-

sages. The best result reached by (Martins et al.,

2021a) had an F1 score of 0.778, considering the full

corpus of COVID-19.BR dataset. In (Martins et al.,

2021b) the authors detailed the dataset proposed in

(Martins et al., 2021a) and presented a case study ex-

ploring data visualization concepts to represent infor-

mation graphically, highlighting patterns and trends in

data and achieving new insights concerning COVID-

19 misinformation on WhatsApp. In (Martins et al.,

2021c), the authors proposed a new approach to mis-

information detection, called MIDeepBR, based on

BiLSTM neural networks, BERT Embeddings, pool-

ing operations and attention mechanisms. MIDeepBR

can automatically detect misinformation in PT-BR

WhatsApp messages. Their best results achieved an

F1 score of 0.834. In (Martins et al., 2022), the

authors explored a posthoc interpretability method

called LIME to explain the predictions of misinfor-

mation detection approaches. Besides, they applied

a textual analysis tool called LIWC to analyze What-

sApp messages’ linguistic characteristics and identify

psychological aspects present in misinformation and

non-misinformation messages. The results indicated

that it is feasible to understand relevant aspects of the

MID model’s predictions and ﬁnd patterns on What-

sApp messages about COVID19.

In (Ng and Loke, 2021), the authors analyzed

a Singapore-based COVID-19 Telegram group with

more than 10000 participants focusing on ﬁve dimen-

sions: participation, sentiment, negative emotions,

topics, and message types. In (J

unior et al., 2022a;

unior et al., 2022b), the authors presented the “Tele-

gram Monitor”, a web-based system that monitors the

political debate in this platform and enables the anal-

ysis of the most shared content in multiple channels

and public groups. In (Paschalides et al., 2020), the

authors presented MANDOLA, a big-data process-

ing system that monitors, detects, visualizes, and re-

ports the spread and penetration of online hate-related

speech using big-data approaches. MANDOLA con-

sists of six individual components that intercommuni-

cate to consume, process, store, and visualize statisti-

cal information regarding hate speech spread online.

ICEIS 2023 - 25th International Conference on Enterprise Information Systems

238

3 THE BATMAN PLATFORM

This section will present the main components of the

BATMAN (Big dATa platforM for misinformAtion

moNitoring) platform, a real-time platform for ﬁnd-

ing, gathering, analyzing, and visualizing misinfor-

mation in social networks, in particular, in instant

message applications such as WhatsApp and Tele-

gram, based on Big Data technologies.

The proposed platform architecture comprises six

layers, as illustrated in Figure 1. Next, we will discuss

in detail each one of these layers and its components.

3.1 Data Collector Layer

The main goal of this layer is make possible gather-

ing data from different social networks using a com-

mon interface. In this layer, the components are called

“connectors”, which are applications that collect data

from a particular social network. The connectors run

in Docker containers and they are independent of each

other. So, a failure in a certain connector does not

affect the others. They collect messages (text) and

media (image, audio and video) shared in a speciﬁc

social network. A connector converts each captured

message to the JSON (JavaScript Object Notation)

format and send it to the Message Broker (Redis)

on the process layer. We chose JSON because it is

a language-independent, standard format for storing

and exchanging data. Besides, a connector sends each

captured media ﬁle to the File Server component on

the Persistence Layer. In order to avoid storing du-

plicate media ﬁles, we apply the MD5 hash algorithm

on the ﬁle content and generate a a unique identiﬁer,

which is used as the name of the media ﬁle. Thus, we

avoid wasting disk space, as well as making it pos-

sible to aggregate similar content and quantify how

many times each one was shared by the users, with the

purpose of understand the popularity of each content.

Currently, the BATMAN platform has two different

connectors running: one for WhatsApp and other to

Telegram. Listings 1 and 2 illustrate examples of

JSON “ﬁles” caught by WhatsApp and Telegram con-

nectors, respectively.

3.2 Process Layer

The main purpose of this layer is ensure a common in-

terface for receiving the messages collect by connec-

tors. This layer has two components: Message Broker

(Redis) and ETL Application.

The Message Broker is a software that make it

possible that applications, systems and services com-

municate with each other and share information. It is

responsible for validating, storing, routing and deliv-

ering messages to the appropriate destinations. It acts

as an intermediary between other applications, allow-

ing senders (connectors, in this case) to issue mes-

sages without knowing where the receivers (ETL Ap-

plication, in this case) are, whether they are active or

not, or how many of them there are. This facilitates

the decoupling of processes and services within the

proposed architecture. The Message Broker allows

reliable storage and ensures message delivery. It has

a set of message queues, which store and sort mes-

sages until consuming applications can process them.

Furthermore, it ensures that each queued message is

consumed only once. To implement the Message Bro-

ker, we use Redis, which is an in-memory, key-value,

open source, versatile and easy-to-use storage system.

In addition, it provides high performance, persistence

and data replication.

The ETL Application is responsible for the mes-

sage processing, which includes different tasks, such

as: parsing, anonymization, user geolocation discov-

ery (only for WhatsApp messages), misinformation

detection and sentiment analysis. Many of these tasks

use the services of the Data Processing API on the

Data Processing Layer. We took into consideration

privacy issues by anonymizing users’ names and cell

phone numbers. For this, we create an anonymous

and unique ID for each user by using an MD5 hash

function on their phone number. Similarly, we create

an anonymous alias for each group. After a message

processing, the ETL Application component sends

the resulting data to a Relational Database Server

(PostgreSQL) and a Search Engine (Elasticsearch),

both on the Persistence Layer.

3.3 Data Processing Layer

The main goal of this layer is to provide a set of

services to support the data processing, through a

standardized API (Application Programming Inter-

face).This API integrates several independent com-

ponents, which will be detailed next. The Machine

Learning Models component includes a service to

compute the probability of a text message received

as input to contain misinformation. The Geographic

Component provides a service to discover the geo-

graphic location (DDD and DDI) of a WhatsApp user.

The Natural Language Processing Component has a

service to compute the sentiment (polarity) of a text

message received as input. The Image Processing

component is under development and will provides

services for extract text from image ﬁles collected

from the connectors, for ﬁnd similar images and for

identify objects in an image.

BATMAN: A Big Data Platform for Misinformation Monitoring

239

Figure 1: The BATMAN Platform Architecture.

3.4 Persistence Layer

The main purpose of this layer is to provide sup-

port for storing and querying data. This layer has

four components: File Server, Search Engine, Re-

lational Database Server and Multi-instance Integra-

tion. Next, we will describe each one of them.

The File Server component is responsible for store

the media ﬁles (audios, images and videos) cap-

tured by the connectors in a persistent and safe man-

ner. The Search Engine component aims to pro-

vide textual queries directly on the captured mes-

sages. For this, it uses Elasticsearch, a search en-

gine based on the Lucene library that provides a dis-

tributed, multitenant-capable full-text search engine

with an HTTP web interface and schema-free JSON

documents. The Relational Database Server sup-

ports storing and querying data on the traditional ﬂat

model. Thereunto, it uses PostgreSQL, a free and

open-source relational database management system

(RDBMS) emphasizing extensibility and SQL com-

pliance. Figure 2 illustrates the PostgreSQL database

schema, that is, its tables and columns. It is important

to highlight that the audios, images, and videos are

stored by the File Server. The PostgreSQL database

stores only the path to these ﬁles. The Multi-instance

Integration component is under development and will

provides support to integrate and communicate sev-

eral distributed instances of the BATMAN platform.

3.5 Data Visualization Layer

The main goal of this layer is to support visualization

and consuming of the data previously stored and pro-

cessed. This layer includes four components: Data

Access API, Web Portal, WhatsApp Bot and Tele-

gram Bot. Next, we will describe each one of these

components.

The Data Access API has a set of services to ac-

cess: i) processed data stored on the relational model

ICEIS 2023 - 25th International Conference on Enterprise Information Systems

240

Listing 1: An Example of a JSON Caught from WhatsApp.

{

” i d m e s s a g e ” : ” EE6CF9B6E75AE22708BDDE4B17548D6D ” ,

” m ess e n g e r ” : ” w h a ts a p p ” ,

” m e s s a g e t y p e ” : ” DocumentoComLegenda ” ,

” i d p e r s o n a ” : ”XXXXXXXXXXXX. 0 : 1 2 @s . w h a t s a pp . n e t ” ,

” d a t e m e s s a g e ” : ”2022 −09 −17 0 1 : 4 1 : 0 5 +0000 UTC” ,

” t e x t c o n t e n t ” : ” ” ,

” i d memb er ” : ”XXXXXXXXXXXXX@s. wh a t s a p p . n e t ” ,

” i d g r o u p ” : ”YYXXZZZZ09789−1513745318@g. us ” ,

” media ” : ”833 f 1 4 a 3 b c b b b 7 f 5 c 4 d 3 5 6 c f e 1 a b 1 9 f a . p d f ” ,

” media name ” : ” P a g i n a s do L u l a f l i x , s i t e com a s f a l c a t r u a s

de L u l l a que a J u s t i c a mandou t i r a r do a r − C o ns e g u i s a l v a r t o d a s a s

p a g i n a s . pd f ” ,

” m e d i a t y p e ” : ” a p p l i c a t i o n / p d f ” ,

” m e d i a u r l ” : ” ” ,

” media md5 ” : ”833 f 1 4 a 3 b c b b b 7 f 5 c 4 d 3 5 6 c f e 1 a b 1 9 f a ” ,

” d i s p l a y n a m e ” : ” ” ,

” a d d r e s s m e s s a g e ” : ” ” ,

” l a t i t u d e m e s s a g e ” : 0 ,

” l o n g i t u d e m e s s a g e ” : 0 ,

” c o n t a c t s m e s s a g e ” : n u l l

}

(PostgreSQL), ii) processed data stored on text for-

mat (Elasticsearch) and iii) media ﬁles previously col-

lected by the connectors. Thus, it uses the Data Pro-

cessing API on the Data Processing Layer. Today,

there is a great need for displaying massive amounts

of data in a way that is easily accessible and under-

standable. In this context, data visualization is a way

to represent information graphically, highlighting pat-

terns and trends in data and helping to achieve news

insights. It enables the data exploration via the ma-

nipulation of charts and images. More speciﬁcally,

it enables users to analyze the data by interacting

directly with a visual representation of it. In this

work, the Web Portal component is a web application

developed using Python programming language and

Django 3 framework, which explores relational (from

PostgreSQL) and textual (from Elasticsearch) data.

The last two components of the Data Visualization

Layer, WhatsApp Bot and Telegram Bot, are proac-

tive chatbots, which automatically detects and alerts

the presence of misinformation in social chats. Ini-

tially, they need to be added to a certain group. Then

it will automatically monitor and analyze the content

that travels in the group. Finally, if they detect that

a certain content has a high probability of containing

misinformation, an alert message is sent to the group.

3.6 Monitoring Management Layer

The main purpose of this layer is monitoring the oper-

ation of the BATMAN platform as a whole and alerts

a human administrator by email and SMS (short mes-

sage system) in case of failures. Besides, this layer

maintains a set of logs, which can be used to audit,

troubleshooting and repairs.

4 CASE STUDY

To evaluate the BATMAN platform, we performed an

exploratory case study using two different datasets,

covering the Brazilian general elections campaign in

2022, collected by WhatsApp and Telegram, respec-

tively. Next, we will describe these two datasets in

detail.

• Brazilian general elections on WhatsApp: This

dataset contains 798,882 messages, obtained from

17,717 users (cell phone chips), which partici-

pated of 331 WhatsApp public groups, in the pe-

riod from August to November 2022.

• Brazilian general elections on Telegram: This

dataset contains 561,449 messages, obtained from

14,866 users, which participated of 180 Telegram

public groups, in the period from September to

November 2022.

BATMAN: A Big Data Platform for Misinformation Monitoring

241

Listing 2: An Example of a JSON Caught from Telegram.

{

” i d m e s s a g e ” : ” 1 0 49 8 1 9 ” ,

” m e s s e n g e r ” : ” t e l e g r a m ” ,

” m e s s a g e t y p e ” : ” U r l ” ,

” i d p e r s o n a ” : ######### ,

” d a t e m e s s a g e ” : ”2022 −09 −28 T16 : 1 8 : 0 9 + 0 0 : 0 0 ” ,

” t e x t c o n t e n t ” : ” Banco M u n d i a l c o n f i r m a f a l a de P a u l o Guedes : PIB do

B r a s i l t e n d e a c r e s c e r m a is que o da China \ u261b h t t p s : / /

t e r r a b r a s i l n o t i c i a s . com / 2 0 2 2 / 0 9 / banco − mundial −c o n f i r m a − f a l a a −de − p au l o

− g u e d e s −pib −do − b r a s i l − t e n d e −a − c r e s c e r − mais −que −o−da − c h i n a / \ n\nBom d i a

\ ud83d \ udd25 \ ud83d \ udd25 \n\n@FYIBRASIL ” ,

” i d m e m b e r t e l e g r a m ” : ######### ,

” i d

g r o u p ” : ##### # # # # ,

” media ” : ” c a 9 a 6 c 2 5f 2 9 5 2 6 9 3 0 d 60 8 52 a 7 4 a b 6 94 0 ( 1 ) . j p g ” ,

” media name ” : ” ” ,

” m e d i a t y p e ” : ” u r l ” ,

” m e d i a u r l ” : ” h t t p s : / / t e r r a b r a s i l n o t i c i a s . com / 2 0 2 2 / 0 9 / banco − m u n d i a l −

c o n f i r m a − f a l a a −de − p a u l o − guedes −pib −do− b r a s i l − t e n d e −a − c r e s c e r − mais −que

−o−da − c h i n a / ” ,

” media md5 ” : ” c a9 a 6c 2 5 f 2 9 5 26 9 30 d 6 0 8 5 2 a 7 4 ab 6 94 0 ” ,

” d i s p l a y n a m e ” : ” Banco Mund i a l c o n f i r m a f a l a de P a u l o Guedes : PIB do

B r a s i l t e n d e a c r e s c e r m a is que o da China − T e r r a B r a s i l Not \

u 0 0 e d c i a s \ n C o m p a r t i l h e : A nova p r e v i s \ u00e3o do Banco Mundi a l \ u00e9

de que o PIB ( P r o d u t o I n t e r n o B r u t o ) ” ,

” a d d r e s s m e s s a g e ” : ” ” ,

” l a t i t u d e m e s s a g e ” : 0 ,

” l o n g i t u d e m e s s a g e ” : 0 ,

” c o n t a c t s m e s s a g e ” : n u l l

}

Using the Web Portal component of the the Data

Visualization layer from the BATMAN Platform, the

user can choose a speciﬁc dataset or all data from all

datasets to build a set of visualizations.

4.1 Messages Characterization

In general, messages created to spread misinforma-

tion include a URL, often from a little-known website

or blog, to give it credibility. Thus, the presence of a

URL can be a criterion for selecting messages to be

analyzed by fact-checkers. We observed that a signif-

icant proportion of the caught Telegram (29.85%) and

WhatsApp (20.16%) messages contains some URL.

Currently, audios, images, and videos are com-

monly used to spread misinformation. Therefore, the

messages associated with these ﬁles are potential can-

didates to undergo a veriﬁcation process. We ob-

served that a signiﬁcant proportion of the caught Tele-

gram (63.32%) and WhatsApp (62.60%) messages

contains some media ﬁle.

Figure 3 shows the distribution messages sending

time by the day hours on Telegram. As we can imag-

ine, the peak of sending messages occurs at the time

reserved for lunch (between 12 and 15 hours) and in

the early evening, just after work hours.

Figure 4 shows the distribution messages send-

ing time by day on Telegram. As we can imagine,

the peak of sending messages occurs on October 2nd

(date of the ﬁrst round of elections) and October 30th

(date of the second round of elections).

4.2 Geographic Distribution

In the 2020 Brazilian elections, some cell phone chips

from foreign countries were used in the electoral ad-

vertisement. Thus, monitor the messages sent by

these chips is an important task to identify misinfor-

mation spreading. We observed that 1,83% of What-

sApp messages were sent by cell phone chips of for-

eign countries.

ICEIS 2023 - 25th International Conference on Enterprise Information Systems

242

Figure 2: The PostgreSQL Database Schema.

Figure 3: Number of Messages by Hour.

Another relevant aspect to observe in the moni-

tored groups is the geographic location of users (cell

phone chips), both Brazilians and foreigners, besides

these users’ activity level. Figure 5 shows the Brazil-

ian states with more quantity of messages on What-

sApp. As might be expected, the most populous states

have the most signiﬁcant amount of messages sent.

Figure 6 illustrates the Brazilian states with more

Figure 4: Number of Messages by day.

users’ on WhatsApp. The most populous states have

the most signiﬁcant amount of users.

However, when analyzing the states with more

messages per user (Figure 7), we can observe that not

so populous states such as Paraiba, Alagoas, and Dis-

trito Federal, have the most active users.

BATMAN: A Big Data Platform for Misinformation Monitoring

243

Figure 5: States with more Messages.

Figure 6: States with more Users.

4.3 Vocabulary Characterization

Another aspect that needs to be analyzed is the char-

acteristics of the vocabulary used in the text messages,

since there is a strong relationship between the used

vocabulary and the social network, in this case, What-

sApp. Figures 8 and 9 show the number of messages

by the number of words contained in the message, for

the WhatsApp and Telegram datasets, respectively.

As we can note, there are few messages with a large

number of words and a high number of messages with

few words.

Figures 10 and 11 show the word cloud highlight-

ing the most popular words on WhatsApp and Tele-

gram, respectively.

Figure 7: States with more Messages per User.

4.4 Misinformation Analysis

The last aspect to be explored using the Web Por-

tal component on the Data Visualization layer of the

BATMAN Platform is the misinformation analysis. In

this context, various information about messages and

users are explored to identify text messages contain-

ing misinformation and super-spreaders, that is, the

users that most spread misinformation.

Table 1 contains the ﬁve most shared messages

on WhatsApp. The “Sharings” column indicates how

many times the message was shared. It is important to

highlight that all the ﬁve most shared messages con-

tain misinformation.

Finally, we can query the URLs most used in

the messages. Tables 2 and 3 contain the ﬁve most

shared URLs together with the number of messages

that refers each URL, on WhatsApp and Telegram,

respectively.

5 CONCLUSIONS

The fast spread of misinformation through social net-

works, such as WhatsApp and Telegram, messages

poses a signiﬁcant social problem. In this work, we

presented BATMAN, a platform for ﬁnding, gather-

ing, analyzing, and visualizing misinformation in so-

cial networks. To evaluate our methodology, we built

two different datasets. Besides, we presented a case

study using the proposed platform. We hope that our

platform can help journalists and researchers to un-

derstand the misinformation propagation in Brazil.

ICEIS 2023 - 25th International Conference on Enterprise Information Systems

244

Figure 8: Number of Messages by the Number of Words

in the Message on WhatsApp.

Figure 9: Number of Messages by the Number of Words

in the Message on Telegram.

Figure 10: Word Cloud on WhatsApp.

Figure 11: Word Cloud on Telegram.

Table 1: Most Shared Messages on WhatsApp.

Sharings Text

275 Pesquisa eleitoral para presidente da Rep

ublica em 2022. Vote https://ﬁnancas-

web.club/enquete/

190 Lula ou Bolsonaro? Quem est

a ganhando? Descubra a verdade e vote

https://eleicoes2022.info/atualizada

173 Pesquisa eleitoral para presidente da Rep

ublica em 2022. Vote

https://www.creditoparacartao.com/enquete/

79 AO VIVO: NOVA ENQUETE OFICIAL para Presidente da Republica 2022: Entre e Vote

==¿¿ https://enqueteeleicoes2022.com.br/

77 Atenc¸

ao, Brasil! Chegou a hora de colocarmos ﬁm aos abusos praticados pelo ministro

Alexandre de Moraes e mostrarmos aos demais ministros do STF que o povo n

ao aceitar

mais qualquer excesso praticado por eles. Como qualquer outro membro do Estado, eles de-

vem agir nos limites da lei e de sua compet

encia. Para isso, vamos fazer circular essa petic¸

pelo IMPEACHMENT de Alexandre de Moraes. O Brasil acordou e vamos mostrar que

supremo

e o POVO! Assine e compartilhe! Pra frente, Brasil! Vamos mostrar nossa forc¸a!

https://peticaopublica.me/impeachment-alexandre-de-moraes-2021/

Table 2: Most Shared URLs on WhatsApp.

Qtd Site

4212 https://h3r0.link/9XLudJ8VU12CYzpJ7

3583 https://api.whatsapp.com/sendphone=CCDDDXXXXXXX %20inter-

esse%20

2473 https://t.me/+SLVkezIiNKkkH4sy

761 https://t.me/apostagem

622 https://www.youtube.com/channel/UCou3uZZFuvu5oB E7BOXf6Q

BATMAN: A Big Data Platform for Misinformation Monitoring

245

Table 3: Most Shared URLs on Telegram

Qtd Site

1221 https://t.me/canalselvabrasiloﬁcial

1188 https://youtu.be/qbTzhB0akt8

1049 https://youtu.be/zDuOoyhyN-4

659 https://youtu.be/4DHk9KZ01HM

593 https://youtu.be/x2uiakywcrI

REFERENCES

Cabral, L., Monteiro, J. M., da Silva, J. W. F., Mattos, C.

L. C., and Mour

ao, P. J. C. (2021). Fakewhastapp.br:

NLP and machine learning techniques for misinfor-

mation detection in brazilian portuguese whatsapp

messages. In Filipe, J., Smialek, M., Brodsky, A.,

and Hammoudi, S., editors, Proceedings of the 23rd

International Conference on Enterprise Information

Systems, ICEIS 2021, Online Streaming, April 26-28,

2021, Volume 1, pages 63–74. SCITEPRESS.

de S

a, I. C., Monteiro, J. M., da Silva, J. W. F., Medeiros,

L. M., Mour

ao, P. J. C., and da Cunha, L. C. C.

(2021). Digital lighthouse: A platform for monitor-

ing public groups in whatsapp. In Filipe, J., Smi-

alek, M., Brodsky, A., and Hammoudi, S., editors,

Proceedings of the 23rd International Conference on

Enterprise Information Systems, ICEIS 2021, Online

Streaming, April 26-28, 2021, Volume 1, pages 297–

304. SCITEPRESS.

Garimella, K. and Tyson, G. (2018). Whatsapp, doc? a ﬁrst

look at whatsapp public group data. arXiv preprint

arXiv:1804.01473.

unior, M., Melo, P. F., Kansaon, D., Mafra, V., S

a, K.,

and Benevenuto, F. (2022a). Telegram monitor: Mon-

itoring brazilian political groups and channels on tele-

gram. CoRR, abs/2202.04737.

unior, M., Melo, P. F., Kansaon, D., Mafra, V., S

a, K., and

Benevenuto, F. (2022b). Telegram monitor: Moni-

toring brazilian political groups and channels on tele-

gram. In Bellog

ın, A., Boratto, L., and Cena, F., edi-

tors, HT ’22: 33rd ACM Conference on Hypertext and

Social Media, Barcelona, Spain, 28 June 2022- 1 July

2022, pages 228–231. ACM.

Machado, C., Kira, B., Narayanan, V., Kollanyi, B., and

Howard, P. (2019). A study of misinformation in

whatsapp groups with a focus on the brazilian presi-

dential elections. WWW ’19, page 1013–1019, New

York, NY, USA. Association for Computing Machin-

ery.

Martins, A. D. F., Cabral, L., Mour

ao, P. J. C., Monteiro,

J. M., and Machado, J. (2021a). Detection of misinfor-

mation about covid-19 in brazilian portuguese what-

sapp messages. In International Conference on Appli-

cations of Natural Language to Information Systems,

pages 199–206. Springer.

Martins, A. D. F., da Cunha, L. C. C., Mour

ao, P. J. C.,

de S

a, I. C., Monteiro, J. M., and de Castro Machado,

J. (2021b). Covid19.br: A dataset of misinformation

about covid-19 in brazilian portuguese whatsapp mes-

sages. In III Dataset Showcase Workshop, DSW 2021,

Rio de Janeiro, RJ, Brazil, October 4-8, 2021 (To ap-

pear). SBC.

Martins, A. D. F., da Cunha, L. C. C., Mour

ao, P. J. C.,

Monteiro, J. M., and de Castro Machado, J. (2021c).

Detection of misinformation about covid-19 in brazil-

ian portuguese whatsapp messages using deep learn-

ing. In XXXVI Simp

osio Brasileiro de Banco de Da-

dos, SBBD 2021, Rio de Janeiro, RJ, Brazil, October

4-8, 2021 (To appear). SBC.

Martins, A. D. F., Monteiro, J. M., and Machado, J. C.

(2022). Understanding misinformation about COVID-

19 in whatsapp messages. In Chiusano, S., Cerquitelli,

T., Wrembel, R., Nørv

ag, K., Catania, B., Vargas-

Solar, G., and Zumpano, E., editors, New Trends

in Database and Information Systems - ADBIS 2022

Short Papers, Doctoral Consortium and Workshops:

DOING, K-GALS, MADEISD, MegaData, SWODCH,

Turin, Italy, September 5-8, 2022, Proceedings, vol-

ume 1652 of Communications in Computer and Infor-

mation Science, pages 14–23. Springer.

Newman, N., Fletcher, R., Schultz, A., Andi, S., and

Nielsen, R. K. (2021). Reuters institute digital news

report 2021. Dados sobre a utilizac¸

ao de redes sociais

no Brasil.

Ng, L. H. X. and Loke, J. Y. (2021). Analyzing public opin-

ion and misinformation in a covid-19 telegram group

chat. IEEE Internet Computing, 25(2):84–91.

Paschalides, D., Stephanidis, D., Andreou, A., Orphanou,

K., Pallis, G., Dikaiakos, M. D., and Markatos, E.

(2020). Mandola: A big-data processing and visu-

alization platform for monitoring and detecting online

hate speech. ACM Trans. Internet Technol., 20(2).

Resende, G., Melo, P., Sousa, H., Messias, J., Vascon-

celos, M., Almeida, J., and Benevenuto, F. (2019).

(mis)information dissemination in whatsapp: Gather-

ing, analyzing and countermeasures.

Resende, G., Messias, J., Silva, M., Almeida, J., Vascon-

celos, M., and Benevenuto, F. (2018). A system for

monitoring public political groups in whatsapp. In

Proceedings of the 24th Brazilian Symposium on Mul-

timedia and the Web, WebMedia ’18, page 387–390,

New York, NY, USA. Association for Computing Ma-

chinery.

Silva, M. and Benevenuto, F. (2021). COVID-19 ads as po-

litical weapon. In Hung, C., Hong, J., Bechini, A., and

Song, E., editors, SAC ’21: The 36th ACM/SIGAPP

Symposium on Applied Computing, Virtual Event, Re-

public of Korea, March 22-26, 2021, pages 1705–

1710. ACM.

ICEIS 2023 - 25th International Conference on Enterprise Information Systems

246