Integrating User Sentiment into Software Evolution Processes

对于拥有大量用户基础和快速发展需求的大型软件应用程序来说，维护其质量是一项挑战。源代码重构、面向对象重构和缺陷预测方法是积极的研究工作的例证，这些研究工作旨在随着应用程序的发展而提高应用程序的质量。然而，应用程序的质量并不总是反映该应用程序的感知质量。具体来说，感知质量捕获了用户对应用程序和实际应用程序的缺陷的感知和情绪（例如，用户的态度、意见或感觉）。例如，一个有30个很少被展示的缺陷的应用程序可能比一个有一个经常被展示的缺陷的应用程序具有更好的感知质量。

It is challenging to maintain the quality of large-scale software applications with a large user base and rapidly evolving requirements. Source code restructuring, object oriented refactoring and defect prediction approaches exemplify active research efforts to improve the quality of applications as they evolve. However, the quality of an application does not always reflect perceived quality of that application. Specifically, perceived quality captures users' perception and sentiment (e.g., users' attitude, opinion or feeling) towards an application and the faultiness of the actual application. For example, an application with 30 defects which are rarely exhibited may have better perceived quality than an application with one defect that is exhibited frequently. Different types of user studies, e.g., direct observation and laboratory methods, are often used to study users' perception of the quality of an application. However, such types of studies are costly to operate, and it is hard to interpret the subjective results across different user groups. As the proliferation of various social channels, such as social media and social networks, has shot up during the past decade, users prefer to provide instant feedback on an application through social channels which are more intuitive for users, instead of submitting an issue report for a defect. For applications with a large user base, a large amount of user feedback can be collected over social channels (e.g., Twitter and Stack Overflow). Such user feedback can uncover a wider range of problematic usage scenarios which occur under natural usage environments. In this proposed research program, we are interested in capturing perceived quality of applications through analyzing the enormous amount of crowdsourced data available across various social channels and leveraging such information to improve perceived quality of applications as they evolve. We focus on two themes: (1) Integrating perceived quality into the defect fixing process to help practitioners prioritize defects with the highest impact on perceived quality and to provide instant developer response to unfavorable user sentiment; and (2) Leveraging perceived quality to allow practitioners to quickly adapt their applications to the evolving software components and help practitioners select components with high perceived quality for software integration. The result of the research can help practitioners focus on high impact issues, thereby increasing customer satisfaction, brand reputation and ultimately company revenues. The research will directly benefit the software evolution processes at some of Canada's top IT companies, e.g., IBM and Blackberry who are generously providing us with access to their datasets and software to conduct our research. The proposed research program will train 6 HQPs (i.e., 3 PhD and 3 MSc) in an area of great importance to Canada's economy.