Measuring the emotional charge: Analysis of the emotions present in the content of tweets about COVID-19 in Lima

Holgado-Apaza L.A.; Ancco-Calloapaza C.L.; Bedregal-Flores O.; Quispe-Layme M.; Miranda-Castillo R. · Revista Cientifica de Sistemas e Informatica (2023) · DOI: 10.51252/rcsi.v3i2.587

Resumen

During the state of emergency and quarantines implemented by world leaders, there has been a significant increase in people's activity on social networks, such as Twitter, where they share opinions and emotionally charged news. In this study, we present a visualization tool for sentiment analysis in tweets related to COVID-19 in the city of Lima, Peru, during the year 2020. For this purpose, we train a BERT model called BETO, specifically designed for natural language processing in Spanish. We used the SenWave dataset, comprising 11 emotions, to train the model. Subsequently, we validate the model using a dataset composed of 33,770 tweets collected in the city of Lima, Peru. The result of our study is an interactive dashboard showing the flow of sentiments expressed in the analyzed tweets. Our findings reveal that the three most frequent emotions during 2020 were: humor, boredom and optimism. In addition, we identified the five most popular words used in the tweets: contagion, health, distancing, isolation and Martín Vizcarra, referring to the former president of Peru.

Instituciones: Universidad Nacional de San Agustín de Arequipa; Universidad Nacional Amazónica de Madre de Dios