issue_classify/abstract.tex

22 lines
2.1 KiB
TeX
Raw Permalink Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

Currently, open source projects receive various kinds of issues daily,
because of the extreme openness of Issue Tracking System (ITS) in GitHub.
ITS is a labor-intensive and time-consuming task of issue categorization for project managers.
%,and utilizing automatic approach to predict category of issue reports can effectively reduce management costs.
%Many researches have worked on this problem in ITS like Bugzilla, ITracker, etc,
%and adding structured information of ITS to classifier is a common method to improve performance of classification model.
However, a contributor is only required a short textual abstract to report an issue in GitHub.
Thus, most traditional classification approaches based on detailed and structured data
(\eg priority, severity, software version and so on) are difficult to adopt.
In addition, mislabeled issues is common in ITS, but less works study how mislabeled data affect the performance of classification model.
%However, nearly all research works on few projects, and we don't know how these methods perform for most projects.
%Otherwise, for ITS of GitHub, these structured information is hard to collect, which makes the method useless.
In this paper, issue classification approaches on a large-scale dataset,
including 80 popular projects and over 252,000 issue reports collected from GitHub, were investigated.
First, four traditional text-based classification methods and their performances were discussed.
Semantic perplexity (\ie an issues description confuses bug-related sentences with
nonbug-related sentences) is a crucial factor that affects the classification performances based on quantitative and qualitative study.
Then, A two-stage classifier framework based on the novel metrics of semantic perplexity of issue reports was designed.
Results show that our two-stage classification can significantly improve issue classification performances.
Finally, we study how mislabeled issues influence performance of issue classification model.
We find when there are many semantic perplexity issues existing in dataset, our 2-stage framework has a well anti-noise ability.