Complete Bug Report Summarization Using Task-Based Evaluation

Abstract

A software project's bug reports provide a rich source of information for software Developer in different tasks like understanding multiple aspects of particular defect when working on the project. For interaction with bug reports developers required some text so, in this topic investigated if it is possible to summarize bug reports automatically so that developers can performs their tasks by deliberating short summaries instead of entire bug report. The proposed system deals with existing conversation-based automated summarizers and found that the quality of generated summaries is similar to summaries produced for email threads and other conversations. It also focused on training a summarizer on a bug report corpus which helps to check summaries that are statistically better than summaries produced by existing conversation-based generators. For bug report duplicate detection tasks, system conduct a task based evaluation so the automatic produced bug reports summaries can help a developer for their tasks and save time of study participant. There was no proofs which show that accuracy is become weaken when summaries were used and that most participants preferred working with summaries to working with original bug reports.