New features for duplicate bug detection

Abstract

Issue tracking software of large software projects receive a large volume of issue reports each day. Each of these issues is typically triaged by hand, a time consuming and error prone task. Additionally, issue reporters lack the necessary understanding to know whether their issue has previously been reported. This leads to issue trackers containing a lot of duplicate reports, adding complexity to the triaging task. Duplicate bug report detection is designed to aid developers by automatically grouping bug reports concerning identical issues. Previous work by Alipour et al. has shown that the textual, categorical, and contextual information of an issue report are effective measures in duplicate bug report detection. In our work, we extend previous work by introducing a range of metrics based on the topic distribution of the issue reports, relying only on data taken directly from bug reports. In particular, we introduce a novel metric that measures the first shared topic between two topic-document distributions. This paper details the evaluation of this group of pair-based metrics with a range of machine learning classifiers, using the same issues used by Alipour et al. We demonstrate that the proposed metrics show a significant improvement over previous work, and conclude that the simple metrics we propose should be considered in future studies on bug report deduplication, as well as for more general natural language processing applications.