Publicação
Comparing sentiment analysis tools on gitHub project discussions
| Resumo: | The context of this work is situated in the rapidly evolving sphere of Natural Language Processing (NLP) within the scope of software engineering, focusing on sentiment analysis in software repositories. Sentiment analysis, a subfield of NLP, provides a potent method to parse, understand, and categorize these sentiments expressed in text. By applying sentiment analysis to software repositories, we can decode developers’ opinions and sentiments, providing key insights into team dynamics, project health, and potential areas of conflict or collaboration. However, the application of sentiment analysis in software engineering comes with its unique set of challenges. Technical jargon, code-specific ambiguities, and the brevity of software-related communications demand tailored NLP tools for effective analysis. The study unfolds in two primary phases. In the initial phase, we embarked on a meticulous investigation into the impacts of expanding the training sets of two prominent sentiment analysis tools, namely, SentiCR and SentiSW. The objective was to delineate the correlation between the size of the training set and the resulting tool performance, thereby revealing any potential enhancements in performance. The subsequent phase of the research encapsulates a practical application of the enhanced tools. We employed these tools to categorize discussions drawn from issue tickets within a varied array of Open-Source projects. These projects span an extensive range, from relatively small repositories to large, well-established repositories, thus providing a rich and diverse sampling ground. |
|---|---|
| Autores principais: | Barboza, Lucas Guedes |
| Assunto: | Natural language processing Sentiment analysis Software engineering Opinion mining |
| Ano: | 2023 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Instituto Politécnico de Bragança |
| Idioma: | inglês |
| Origem: | Biblioteca Digital do IPB |
| Resumo: | The context of this work is situated in the rapidly evolving sphere of Natural Language Processing (NLP) within the scope of software engineering, focusing on sentiment analysis in software repositories. Sentiment analysis, a subfield of NLP, provides a potent method to parse, understand, and categorize these sentiments expressed in text. By applying sentiment analysis to software repositories, we can decode developers’ opinions and sentiments, providing key insights into team dynamics, project health, and potential areas of conflict or collaboration. However, the application of sentiment analysis in software engineering comes with its unique set of challenges. Technical jargon, code-specific ambiguities, and the brevity of software-related communications demand tailored NLP tools for effective analysis. The study unfolds in two primary phases. In the initial phase, we embarked on a meticulous investigation into the impacts of expanding the training sets of two prominent sentiment analysis tools, namely, SentiCR and SentiSW. The objective was to delineate the correlation between the size of the training set and the resulting tool performance, thereby revealing any potential enhancements in performance. The subsequent phase of the research encapsulates a practical application of the enhanced tools. We employed these tools to categorize discussions drawn from issue tickets within a varied array of Open-Source projects. These projects span an extensive range, from relatively small repositories to large, well-established repositories, thus providing a rich and diverse sampling ground. |
|---|