A topic-based approach for narrowing the search space of buggy files from a bug report

Abstract

Locating buggy code is a time-consuming task in software development. Given a new bug report, developers must search through a large number of files in a project to locate buggy code. We propose BugScout, an automated approach to help developers reduce such efforts by narrowing the search space of buggy files when they are assigned to address a bug report. BugScout assumes that the textual contents of a bug report and that of its corresponding source code share some technical aspects of the system which can be used for locating buggy source files given a new bug report. We develop a specialized topic model that represents those technical aspects as topics in the textual contents of bug reports and source files, and correlates bug reports and corresponding buggy files via their shared topics. Our evaluation shows that BugScout can recommend buggy files correctly up to 45% of the cases with a recommended ranked list of 10 files.

Learning to rank relevant files for bug reports using domain knowledge

“…These include approaches based on LDA [25,34,40], Latent Semantic Indexing (LSI) [40,45], Smoothed Unigram Model (SUM) [40,45], and SVMs [19,34]. Since BugLocator was reported to outperform the existing approaches using LDA, LSI, and SUM [45], and since BugScout was reported to outperform the SVM model proposed in [34], we expect our LR system to compare favorably with all these previous approaches.…”

“…Table 2 shows the Accuracy@k results of the BugScout (BS) and the LR method, for k = 1, 10, and 20. The BugScout results are copied from [34] and are substantially lower than the LR results. While it is possible that our replicated dataset is different from their dataset, we believe there are two main reasons why the LR method performs better than BugScout:…”

“…1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Compared to BugLocator, BugScout is a more complex approach and thus more difficult to implement correctly. Since we were unable to get access to either the tool or the actual dataset used in [34] , we tried to recreate their dataset by following the description from [12,34]. Thus, we created a test dataset by collecting the specified number of fixed bug reports from the Eclipse Platform (4,136) and AspectJ (271) projects, going backwards from the end of 2010 (when BogScout was published).…”

“…BugScout [34] classifies source files as relevant or not based on an extension to Latent Dirichlet Allocation (LDA) [5].…”

Understanding Review Expertise of Developers: A Reviewer Recommendation Approach Based on Latent Dirichlet Allocation

“…Latent Dirichlet allocation (LDA) is a generative statistical model [15]. Previous studies on text analysis [16][17][18] used LDA to extract topic distribution from textual data.…”