CAREER: Enabling and Exploiting Evidence-Based Bug Triage

开发软件并不是一件简单的任务：在无数的bug报告、功能请求和项目计划中，软件团队必须决定哪些问题值得团队花费有限的时间和资源。为了做出这些决定，大多数团队会进行错误分类，比较对问题频率和严重性的主观估计。这项研究试图用大规模的软件使用数据来取代这些主观估计。为了实现这一点，本研究探索了一种工具，通过人们正常使用帮助工具来检测软件问题。这些工具的使用将以一致、结构化的形式捕捉广泛的软件问题，从而实现多种形式的汇总和分析。与自愿反馈不同，这些工具将成为用户正常工作的一部分，增加频率和严重性估计的代表性，同时也捕获新类型的问题，如非致命错误和广泛的可用性问题。评估包括部署到真实的软件开发团队，并对所得到的工具的有效性进行受控实验。这项工作的更广泛的影响是最终使软件团队能够以最好地满足用户需求的方式发展软件。

Evolving software is no simple task: somehow, amongst innumerable bug reports, feature requests, and project plans, software teams must decide which of issues deserve the team's limited time and resources. To make these decisions, most teams engage in bug triage, comparing subjective estimates of the frequency and severity of issues.This research seeks to replace these subjective estimates with large-scale data on software use. To achieve this, the research explores a tools that detect software issues through peoples' normal use of help tools. The use of these tools will capture a wide range of software issues in a consistent, structured form, enabling several forms of aggregation and analysis. Unlike voluntary feedback, these tools will be part of users' normal work, increasing the representativeness of frequency and severity estimates, while also capturing new kinds of issues such as non-fatal errors and a wide range of usability problems. Evaluation includes deployment to real software development teams and controlled experiments of the efficacy of the resulting tools. The broader impacts of the work are ultimately to enable software teams to evolve software in a way that best meets the needs of its users.