Uma investigação sobre análise de sentimentos e categorização de Issues reabertas do GitHub.

Abstract

The reopening of issues represents a significant challenge in software development and maintenance, increasing the costs and complexity of the efforts involved. This occurrence often indicates unresolved or misunderstood issues in communication between project collaborators and users on platforms like GitHub. This thesis aims to deepen the understanding of issue reopenings in open-source GitHub software repositories, considering historical data, issue categorization, and sentiment analysis of developers involved in the associated discussions. Our methodology involved using the SentiStrength-SE tool, adapted for lexicons in the field of Software Engineering, to calculate polarity and sentiment in the texts of discussions related to issues. Subsequently, we developed an automated issue categorization model, classifying them into specific categories such as configuration, database-related, program anomaly, performance, functional, GUI-related, info, permission/deprecation, network, security, and testing. This approach enables more effective prioritization in resolving reopened issues, directing resources more accurately. Finally, we characterized issue reopenings based on the sentiments of developers expressed in discussions within each issue category. The results revealed that sentiment analysis, when applied in isolation, did not prove to be an effective metric for identifying issue reopenings. However, we identified that certain types of issue categories are more prone to problems related to reopening. This underscores the importance of combining issue categorization with sentiment analysis for a more efficient approach to preventing and addressing issue reopenings in open-source software repositories.