EKD-BSP: Bug Report Severity Prediction by Extracting Keywords from Description
Yanxin Jia, Xiang Chen, Shuyuan Xu, Guang Yang, Jinxin Cao · 2021 8th International Conference on Dependable Systems and Their Applications (DSA) · 2021
Bug severity is important for triagers. Recently, the text in the summary field (i.e., bug summary) of bug reports is usually used to extract features, and then bug report severity prediction models are constructed. In some bug reports, the bug summary may not contain enough useful information. While the text field in the description (i.e., bug description) of bug reports contains detailed information of the bug (e.g., steps to reproduce the bug, stack traces, and expected behavior). However, the bug description may contain irrelevant information. Motivated by the above findings, we propose a novel method EKD-BSP (Bug Report Severity Prediction by Extracting Keywords from Description), which uses the bug summary and the keywords extracted from the bug description to perform severity prediction. Our empirical study selects two large-scale open-source projects (i.e., Eclipse and Mozilla) as the empirical subjects. The empirical results show that EKD-BSP can improve the performance of F-measure by up to 5.19% after compared with the baselines.