Bug Triaging: A Comprehensive Review of Current Practices and the Path Ahead

Bug triaging is important for maintaining software because it helps assign bug reports to the right developers. As software systems grow more complex, traditional methods of triaging can result in mistakes, especially in large projects like Eclipse and Mozilla. Automated bug triaging has made progress in recent years in meeting these challenges. Initially, these methods used text mining and basic machine learning techniques like Naïve Bayes and SVM. Over time, more advanced methods have emerged, including ensemble learning, graph-based models, and hybrid techniques that use text, behavior, and historical data. The use of deep learning models such as CNNs, LSTMs, and BERT has greatly improved the accuracy of triaging by identifying complex patterns in bug reports. Despite these improvements, some issues still exist, such as high computer costs, poor-quality bug data, uneven workloads, and models that struggle to generalize. This paper reviews advances made from 2011 to 2025, highlights current trends, and suggests future directions for automated bug triaging. To our knowledge, no other study has systematically summarized current methods, limits, and research opportunities in this area. This review aims to create a solid foundation for future work, emphasizing the need for scalable, understandable, and flexible solutions in real-world software development.