Mark: 110/110 cum laude

Thesis: “Issue report classification using BERT”

Abstract: “This thesis explores the automatic classification of issue reports in software development, aiming to improve efficiency in issue management and prioritization. It addresses the challenge of misclassification between bug reports, enhancement requests, and questions, which consumes valuable developer resources. Leveraging recent advances in natural language processing, specifically BERT and its variants, the study develops and evaluates supervised learning approaches using issue text and authorship information. The proposed models surpass baseline performance, achieving an F1 score of .8591 using textual data alone. The research includes a comprehensive error analysis to identify key factors influencing classification accuracy. The findings underscore the practical utility of automated issue labeling systems in optimizing software development processes.”

Mark: 110/110 cum laude

Thesis: “Design and development of a classifier for the detection of noninformative comments.”

Abstract: “This thesis work aims to address the problem of identifying non-informative comments within a dataset of comments from open source repositories. The goal is to develop a classifier that can distinguish between informative and non-informative comments in order to support source code analysis and maintenance.”