Toward an Effective Bug Triage System Using Transformers to Add New Developers

As defects become more widespread in software development and advancement, bug triaging has imperative for testing maintenance. The triage process assigns an appropriate developer to a report. Many automated semiautomated systems have been proposed the last decade, some recent techniques provided direction developing effective system. However, these still require improvement. Another open challenge related this problem is adding new developers existing system, which challenging because no listed history. This paper proposes transformer-based system that uses bidirectional encoder representation from transformers (BERT) word representation. model can add without building training scratch. To developers, we assumed had history created by manual triager or human manager after learning their skills Then, was fine-tuned using Experiments were conducted datasets well-known large-scale open-source projects, such as Eclipse Mozilla, top-k accuracy used criterion assessment. experimental outcome suggests better than other word-embedding-based methods problem. Additionally, method performs best requiring retraining whole dataset.