CAREER: Visual Database Interfaces

随着对数据的访问变得越来越普遍，越来越多的人希望从分析数据的能力中受益。不幸的是，占主导地位的数据库接口？SQL？是非常复杂和难以接近的。设计和开发易于理解和高效使用的自定义交互表单和可视化界面是可能的，但是，除了最广泛(通用)或最有价值的任务外，它们是无法实现的，因为构建它们需要相当多的时间和资源。不幸的是，这忽略了一长串简单的分析，这些分析可能永远不会看到数百万用户，但对于取得科学进步、实现个人目标和抓住商业机会仍然至关重要。因此，需要大幅降低构建新接口的成本。拟议的研究将开发可扩展的技术，以自动生成交互式可视化分析界面，这些界面是根据用户的S分析需求和偏好而定制的。这项研究可以对数据驱动的领域产生革命性的影响，如地球科学或生物信息学，并有助于使数据的获取民主化。最后，对数据接口的关注自然适合作为向学生和公众介绍数据分析的教育工具。该项目开发了可扩展的技术来生成可视化数据库接口(VDI)，每个VDI都是为特定的分析任务量身定做的。用户可以使用更便于非技术用户访问、更高效地执行任务并减少用户对技术专家的依赖的VDI，而不是遵循由SQL或通用数据库探索工具定义的界面。为了引导界面生成过程，PI建议利用用户以前尝试过的现有分析查询和工作流的日志。利用这些原木，PI将开发两个新的和互补的系统。第一系统可伸缩地从查询日志中提取分析，并自动生成针对每个分析中的查询定制的交互分析界面。第二种是人在回路界面设计工具，帮助设计人员进一步改进VDIS的交互设计，同时考虑到确保交互具有响应性所需的优化和系统修改。它通过弥合影响最终用户S体验的交互设计需求与影响资源需求的数据结构和系统架构决策之间的鸿沟。该项目将根据用户完成分析任务的能力来评估VDIS，而不是使用传统的或手动设计的界面。PI还将评估大学数据科学课程中的VDI，并与软件一起公开分享课程。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

As access to data becomes more widespread, more and more people want to benefit from the ability to analyze data. Unfortunately, the dominant database interface ? SQL ? is very complex and inaccessible. It is possible to design and develop custom interactive form and visualization interfaces that are easy to understand and efficient to use, however, they are out of reach for all but the most widespread (general-purpose) or highest value tasks due to the considerable time and resources to build them. Unfortunately, this overlooks a long tail of simple analyses that may never see millions of users but are still crucial towards making scientific progress, reaching individual goals, and seizing commercial opportunities. Thus, there is a need to drastically reduce the costs to build new interfaces. The proposed research will develop scalable techniques to automatically generate interactive visual analysis interfaces that are customized to the user?s analysis needs and preferences. This research can have a transformative impact on data-driven fields such as earth sciences or bioinformatics, and help democratize access to data. Finally, the focus on data interfaces is naturally suited as an educational tool to introduce data analysis to students and the public.This project develops scalable techniques to generate Visual Database Interfaces (VDIs), each tailored to specific analysis tasks. Rather than conform to the interface defined by SQL or a generic database exploration tool, users can use VDIs that are more accessible to non-technical users, more efficient for performing tasks, and reduces user reliance on technical experts. To bootstrap the interface generation process, the PI proposes to leverage logs of existing analysis queries and workflows that users have previously attempted. Using these logs, the PI will develop two novel and complementary systems. The first system scalably extracts analyses from query logs, and automatically generates interactive analysis interfaces that are customized to the queries in each analysis. The second is a human-in-the-loop interface design tool that helps designers further improve the interaction design of VDIs with consideration of the optimizations and systems modifications need to ensure the interactions are responsive. It does so by bridging the divide between interaction design requirements that affect the end-user?s experience, and data structure and system architecture decisions that affect resource requirements. The project will evaluate VDIs based on the ability for users to complete analysis tasks as compared to using legacy or manually designed interfaces. The PI will also evaluate VDIs in university courses on Data Science, and share the curricula publicly along with the software.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.