Rrezarta Krasniqi

Rrezarta Krasniqi

In practice, developers search for related earlier bugs and their associated discussion threads when faced with a new bug to repair. Typically, these discussion threads consist of comments and even bug-fixing comments intended to capture clues for facilitating the investigation and root cause of a new bug report. Over time, these discussions can become extensively lengthy and difficult to understand. Inevitably, these discussion threads lead to instances where bug-fixing comments intermingle with seemingly-unrelated comments. This task, however, poses further challenges when dealing with high volumes of bug reports. Large software systems are plagued by thousands of bug reports daily. Hence, it becomes time-consuming to investigate these bug reports efficiently. To address this gap, this paper builds a ranked-based automated tool that we refer it to as RETRORANK. Specifically, RETRORANK recommends bug-fixing comments from issue tracking discussion threads in the context of user query relevance, the use of positive language, and semantic relevance among comments. By using a combination of Vector Space Model (VSM), Sentiment Analysis (SA), and the TextRank Model (TR) we show how that past fixed bugs and their associated bug-fixing comments with relatively positive sentiments can semantically connect to investigate the root cause of a new bug. We evaluated our approach via a synthetic study and a user study. Results indicate that RETRORANK significantly improved performance when compared to the baseline VSM.

Rrezarta Krasniqi

When developers investigate a new bug report, they search for similar previously fixed bug reports and discussion threads attached to them. These discussion threads convey important information about the behavior of the bug including relevant bug-fixing comments. Oftentimes, these discussion threads become extensively lengthy due to the severity of the reported bug. This adds another layer of complexity, especially if relevant bug-fixing comments intermingle with seemingly unrelated comments. To manually detect these relevant comments among various cross-cutting discussion threads can become a daunting task when dealing with high volume of bug reports. To automate this process, our focus is to initially extract and detect comments in the context of query relevance, the use of positive language, and semantic relevance. Then, we merge these comments in the form of a summary for easy understanding. Specifically, we combine Sentiment Analysis and the TextRank Model with the baseline Vector Space Model (VSM). Preliminary findings indicate that bug-fixing comments tend to be positive and there exists a semantic relevance with comments from other cross-cutting discussion threads. The results also indicate that our combined approach improves overall ranking performance against the baseline VSM.