Improving Bug Reporting, Duplicate Detection, and Localization

Abstract: Software developers rely on essential textual information from bug reports (such as Observed Behavior, Expected Behavior, and Steps to Reproduce) to triage and fix software bugs. Unfortunately, while relevant and useful, this information is often missing, incomplete, superficial, ambiguous, or complex to follow. Low-quality content in bug reports causes delay and extra effort on bug triage and fixing. Current technology and research are insufficient to support users and developers on providing high-quality content in bug reports. Our research is intended to fill in this gap, as it aims at improving: (1) the quality of natural language content in bug reports, and (2) the accuracy of Text Retrieval (TR)-based bug localization and duplicate detection. To achieve such goals, our research will identify, enforce, and leverage the discourse that reporters use to describe software bugs.