SHF: Medium: Automated Software Engineering Techniques for Improving the Accessibility of Software

轻松使用软件的能力对每个人都很重要，特别是对世界上大约15%的残疾人来说。即使是普通用户认为理所当然的最简单的操作，对于残疾用户来说也可能是令人生畏的任务。不幸的是，在目前的情况下，软件不可访问性很普遍。这至少可以部分归因于软件工程师可用的现有技术和工具的不足。用于验证软件可访问性的自动化解决方案严重不足。它们要么无法检测到许多真正的可访问性问题，要么报告太多与实践无关的表面问题。自动修复技术，对于提高软件的各种质量属性（例如，可靠性，安全性）是非常有效的，但是对于可访问性来说是稀缺的。现有的交互方式过于死板和繁琐，严重阻碍了残疾用户使用和享受软件技术的深刻进步。该项目为创新技术奠定了基础，使视力和运动障碍的最终用户能够更有效地与软件进行交互。结合经验数据驱动的研究和工具构建活动，该项目将在几个方面推进最先进的技术。首先，研究团队将设计一个用例和辅助服务驱动的可访问性问题检测技术，能够自动识别使用现有技术无法检测到的可访问性问题。其次，研究人员将开发一种自动程序修复解决方案，该解决方案结合了新颖的深度学习和基于搜索的策略来修复各种可访问性问题。此外，团队将构建用于自动识别导航优化用例宏的方法，允许残疾用户通过直观的命令快速执行经常访问的用例。最后，利用用户研究的混合方法以及与残疾用户和软件工程师的访谈，研究人员将评估本项目中开发的技术的功效。最终，该项目将产生一套工具，这些工具将公开提供，以帮助开发人员改进他们构建的软件系统的可访问性。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The ability to use software with ease is important for everyone, especially for approximately 15% of the world population with disabilities. Even the simplest operations, taken for granted by regular users, can be daunting tasks for disabled users. Unfortunately, in the current state of affairs, software inaccessibility is widespread. This can be attributed, at least partly, to the deficiencies in existing techniques and tools available to software engineers. Automated solutions for validating the accessibility of software are woefully insufficient. They either fail to detect many real accessibility issues or report too many superficial issues that are irrelevant in practice. Automated repair techniques, shown to be quite effective for improving various quality attributes of software (e.g., reliability, security), are scarce for accessibility. Existing interaction modalities are too rigid and cumbersome, seriously hindering the disabled users' use and enjoyment of the profound advances in software technology. This project lays the groundwork for innovative technologies that will enable end-users with vision and motor impairments to interact more effectively with software. Combining empirical data-driven research and tool building activities, the project will advance the state-of-the-art in several ways. First, the team of researchers will devise a use-case and assistive-service-driven accessibility issue detection technique capable of automatically identifying accessibility issues that are not detectable using the existing state-of-the-art techniques. Second, the researchers will develop an automated program repair solution that employs a combination of novel deep-learning and search-based strategies for fixing a variety of accessibility issues. Furthermore, the team will construct the means for automatically identifying use case macros for navigation optimization, allowing a disabled user to rapidly execute frequently accessed use cases through intuitive commands. Finally, utilizing a mixed-methods approach of user studies and interviews with both disabled users and software engineers, researchers will evaluate the efficacy of techniques developed in this project. Ultimately, the project will result in a suite of tools, which will be made available publicly, for helping developers with improving the accessibility of software systems that they construct.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.