A carregar...
Publicação
Data Mining e Twitter na gestão da informação em tempos de pandemia no Brasil
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 3.16 MB | Adobe PDF |
Orientador(es)
Resumo(s)
A grande quantidade e crescimento de dados produzidos em redes sociais
proporciona um empoderamento aos seus usuários. A partir desta realidade e a
considerar o contexto pandémico brasileiro, este trabalho aplica técnicas
analíticas de análise de redes sociais, nomeadamente o Twitter. Estas análises
permitem entender as preocupações, as perceções dos usuários e identificar os
sentimentos expressos por usuários desta rede social em relação à pandemia
do Covid-19. Dos 138.648 tweets coletados e analisados do Brasil, entre janeiro
de 2020 a março de 2021, os resultados identificaram que as postagens,
utilizaram palavras ou ideias que remetem principalmente a medidas profiláticas,
para evitar a propagação da doença, como a testagem, as vacinas e aos
impactos sociais e econômicos do progressivo fechamento da economia. O
conjunto de wordclouds, geradas mensalmente, revelou palavras, que dão
indícios dos principais fatos para o período no Brasil e no mundo. Também
ganham destaques, postagens envolvendo as consequências emocionais e
psicológicas do isolamento social, apontando maioria de comentários
classificados como negativos. Os dados demonstraram um comportamento
crescente similar entre a curva acumulada de casos confirmados de Covid-19 e
o número de menções as palavras-chaves adotadas para extração dos tweets.
A análise das tendências das principais palavras empregadas (> 1000 menções
durante todo o período de análise) revelou como mais citadas: “vacina”, com
8.037 menções; “casa”, com 6.816; “Deus”, com 5.130; “contra”, com 5.115;
“caso”, com 4.920; “mortes”, com 4.617; “teste”, com 3.410; “Bolsonaro”, com
2.868 e “tratamento”, com 1.082 citações. Os dados permitiram perceber a
preocupação com o modo como é feita a gestão da pandemia e do “tratamento”
da doença, suas consequências, por um lado, mas também a necessidade de
algo que vai além da ajuda científica, o contraditório e a importância das ações
do presidente. A análise de sentimentos apresentou a polaridade negativa (50%)
predominantemente, seguida da polaridade neutra com 33%. Já para a análise
geográfica, apenas 6% dos tweets apresentavam geolocalização. A Região
Sudeste apresentou o maior número de tweets postados (49%) a maioria deles foi classificado também com a polaridade negativa (56%). Os resultados obtidos
(64%) de acurácia apontam potencialidades do uso dessas técnicas para
análises do Twitter em língua portuguesa. A relação evidente entre tweets e
acontecimentos/comentários revelados por organismos responsáveis ou outros
meios de comunicação comprovam a importância, que a comunicação de risco
pode ter no impacto da comunidade e como uma análise semelhante à efetuada
neste estudo poderá contribuir para a tomada de decisões e avaliação do
impacte de uma comunicação.
The large amount and growth of data produced by users on social networks provides an empowerment to its users. Based on this reality and considering the Brazilian pandemic context, this work applies analytical techniques of social network analysis, namely Twitter. These analyses allow us to understand the concerns, the perceptions of users, and to identify the feelings expressed by users of this social network regarding the Covid-19 pandemic. The 138,648 tweets collected and analyzed from Brazil, between January 2020 to March 2021, identified that the tweets posted, used words or ideas that refer mainly to prophylactic measures, to prevent the spread of the disease, such as testing, vaccines, and to the social and economic impacts of the progressive shutdown of the economy. The set of wordclouds, generated monthly, revealed words, which give indications of the main facts for the period in Brazil and in the world. Posts involving the emotional and psychological consequences of social isolation also stand out, pointing to a majority of comments classified as negative. The data show a similar increasing behavior between the cumulative curve of confirmed cases of Covid-19 and the number of mentions of the keywords adopted to extract the tweets in this work. The trend analysis of the main words used (> 1000 mentions through out the analysis period) revealed as the most cited: "vaccine", with 8,037 mentions; "home", with 6,816; "God", with 5,130; "against", with 5,115; "case", with 4,920; "deaths", with 4,617; "test", with 3,410; "Bolsonaro", with 2,868 and "treatment", with 1,082 citations revealing once again the concern about the way in which the management of the pandemic and the "treatment" of the disease is carried out, its consequences, on the one hand, but also the need for something that goes beyond scientific aid, the contradictory and the importance of the president's actions, on the other. The sentiment analysis showed the negative polarity (50%) predominantly, followed by the neutral polarity with 33%. For the geographic analysis, only 6% of the tweets presented geolocation. The Southeast Region presented the highest number of posted tweets (49%) and most of them were also classified with negative polarity (56%). The obtained accuracy results (64%) point to the potential of using these techniques for Twitter analysis in Portuguese language. The evident relationship between tweets and events/comments revealed by responsible bodies or other media prove the importance that risk communication can have on community impact and how an analysis similar to the one carried out in this study can contribute to decision making and evaluation of the impact of a communication.
The large amount and growth of data produced by users on social networks provides an empowerment to its users. Based on this reality and considering the Brazilian pandemic context, this work applies analytical techniques of social network analysis, namely Twitter. These analyses allow us to understand the concerns, the perceptions of users, and to identify the feelings expressed by users of this social network regarding the Covid-19 pandemic. The 138,648 tweets collected and analyzed from Brazil, between January 2020 to March 2021, identified that the tweets posted, used words or ideas that refer mainly to prophylactic measures, to prevent the spread of the disease, such as testing, vaccines, and to the social and economic impacts of the progressive shutdown of the economy. The set of wordclouds, generated monthly, revealed words, which give indications of the main facts for the period in Brazil and in the world. Posts involving the emotional and psychological consequences of social isolation also stand out, pointing to a majority of comments classified as negative. The data show a similar increasing behavior between the cumulative curve of confirmed cases of Covid-19 and the number of mentions of the keywords adopted to extract the tweets in this work. The trend analysis of the main words used (> 1000 mentions through out the analysis period) revealed as the most cited: "vaccine", with 8,037 mentions; "home", with 6,816; "God", with 5,130; "against", with 5,115; "case", with 4,920; "deaths", with 4,617; "test", with 3,410; "Bolsonaro", with 2,868 and "treatment", with 1,082 citations revealing once again the concern about the way in which the management of the pandemic and the "treatment" of the disease is carried out, its consequences, on the one hand, but also the need for something that goes beyond scientific aid, the contradictory and the importance of the president's actions, on the other. The sentiment analysis showed the negative polarity (50%) predominantly, followed by the neutral polarity with 33%. For the geographic analysis, only 6% of the tweets presented geolocation. The Southeast Region presented the highest number of posted tweets (49%) and most of them were also classified with negative polarity (56%). The obtained accuracy results (64%) point to the potential of using these techniques for Twitter analysis in Portuguese language. The evident relationship between tweets and events/comments revealed by responsible bodies or other media prove the importance that risk communication can have on community impact and how an analysis similar to the one carried out in this study can contribute to decision making and evaluation of the impact of a communication.
Descrição
Dissertação de mestrado, Cultura Científica e Divulgação das Ciências, Universidade de Lisboa, Faculdade de Ciências, Instituto de Ciências Sociais, Instituto de Educação, 2021
Palavras-chave
Gestão da informação - Brasil Covid 19 Pandemia Redes sociais Teses de mestrado - 2021