Mining Issue Trackers: Concepts and Techniques

TL;DR

This paper discusses the challenges of managing complex issue trackers in software development and explores automated natural language processing techniques to analyze issue data for better stakeholder support.

Contribution

It provides a comprehensive overview of concepts and techniques for mining issue trackers, including four major use cases and a practical demonstration package.

Findings

Automated techniques can help manage issue tracker complexity

Natural language processing enables analysis of textual issue data

Four key use cases demonstrate practical applications

Abstract

An issue tracker is a software tool used by organisations to interact with users and manage various aspects of the software development lifecycle. With the rise of agile methodologies, issue trackers have become popular in open and closed-source settings alike. Internal and external stakeholders report, manage, and discuss "issues", which represent different information such as requirements and maintenance tasks. Issue trackers can quickly become complex ecosystems, with dozens of projects, hundreds of users, thousands of issues, and often millions of issue evolutions. Finding and understanding the relevant issues for the task at hand and keeping an overview becomes difficult with time. Moreover, managing issue workflows for diverse projects becomes more difficult as organisations grow, and more stakeholders get involved. To help address these difficulties, software and requirements engineering research have suggested automated techniques based on mining issue tracking data. Given the vast amount of textual data in issue trackers, many of these techniques leverage natural language processing. This chapter discusses four major use cases for algorithmically analysing issue data to assist stakeholders with the complexity and heterogeneity of information in issue trackers. The chapter is accompanied by a follow-along demonstration package with JupyterNotebooks.