Technical Debt

Technical debt, representing suboptimal code choices made for faster delivery, is increasingly addressed through automated detection and management. Current research focuses on leveraging machine learning, particularly transformer-based models like BERT, and ensemble methods to identify self-admitted technical debt (SATD) across various software artifacts (comments, commit messages, issue trackers). This work aims to improve code quality, reduce maintenance costs, and enhance software development efficiency by enabling more accurate and efficient identification and prioritization of debt repayment efforts. The development of larger, more contextually rich datasets and improved model generalizability are key challenges and areas of ongoing investigation.