Análise de sentimentos em commits de projetos de código-fonte aberto no contexto da pandemia de Covid-19

Abstract

The study presented in this research begins with the analysis of feelings expressed in messages written by software developers through commits in open source repositories, covering the years 2019 to 2021. This period encompasses the previous period, during and after the first major waves of the COVID-19 pandemic. Seven open source projects were examined, totaling 173,993 commit messages. The analysis was conducted using two Natural Language Processing (NLP) models based on the RoBERTa architecture, which categorize commit messages into three groups: negative, positive and neutral. Subsequently, an investigation was carried out into the relationship between the feelings expressed in commit messages and the years in which they were written, as well as the effectiveness of each model. The results of this study suggest that messages tend to be more negative at the beginning of COVID-19 waves. After the pandemic, this number tends to decrease. On the other hand, there was an increase in positive messages during this period.