III: Small: One Size Does Not Fit All: Empowering the User with User-Driven Integration

数据和知识集成是代价高昂的过程。因此，大多数现有的解决方案依赖于一种放之四海而皆准的方法，即预先集成数据，然后按原样使用集成的数据或知识库。然而，当数据源是自治的和动态的，或者在大多数科学和决策应用程序中，领域专家的假设、信念和知识对于集成过程是不可或缺的时，这种基于快照的集成解决方案就不能有效地应用。建议的工作解决了用户驱动集成（UDI）系统底层的计算挑战，记住了技术考虑基础上的人为约束和挑战。关键的技术和智力影响在于算法和数据结构，这些算法和数据结构可以通过基于个人用户反馈的用户驱动集成过程，帮助弥合专家用户和系统之间的语义差距。该团队将专门研究(a)通过向量空间嵌入、概率和生成模型不断修订的数据/元数据对齐；(b)查询处理和候选枚举算法，以支持对基于图的数据模型的反馈，并提供不同的解释。UDI在许多领域（比如科学和商业智能）都有潜在的应用，这些领域需要用户驱动的集成来回答不同数据集上的关键问题。特别是，UDI将被纳入美国国家科学基金会资助的tDAR（数字考古记录），这有可能改变考古学。通过极大地提高合成研究的能力来促进美国的科学努力。因此，对基础信息集成挑战的研究将极大地促进科学的共享基础设施，并将使有关复杂系统的重要跨学科研究成为可能。参与这项研究的计算机科学研究生将准备他们在多学科团队中有效地发挥作用，并提高他们对相关挑战和机遇的欣赏。利用UDI作为试验台将使这些学生能够在科学信息管理方面进行实验，从而提高他们对数据综合和科学信息学问题的认识。我们希望有两门研究生课程可以利用数据集和项目软件作为教育平台。亚利桑那州立大学还通过全国认可的住宿荣誉学院和少数民族研究职业计划招募高素质的本科生，该项目将包括本科生荣誉学生参与该项目。UDI还将通过Capstone项目作为本科生的测试平台。

Data and knowledge integration are costly processes. Consequently, most existing solutions rely on a one-size-fits-all approach, where the data are integrated upfront and then the integrated data or knowledge-bases are used as is. Such snapshot-based integration solutions, however, cannot be effectively applied when the data sources are autonomous and dynamic or when, as in most scientific and decision making applications, assumptions, beliefs, and knowledge of the domain experts are indispensable to the integration process.The proposed work tackles the computational challenges underlying a user driven integration (UDI) system, keeping in mind the human constraints and challenges that underlie the technical considerations. The key technical and intellectual impacts are in algorithms and data structures that can help bridge the semantic gap between the expert user and the system through a user-driven integration process based on individual user feedback. The team will specifically investigate (a) continuously revisable data/metadata alignment through vector space embeddings and probabilistic and generative models and (b) algorithms for query processing and candidate enumeration to support feedback over graph-based models of data with alternative interpretations.UDI has potential applications to many domains (such as science and business intelligence) that need user-driven integration to answer key questions over diverse data sets. In particular,UDI will be incorporated into the NSF-funded tDAR (the Digital Archaeological Record), which has the potential to transform archaeology?s scientific endeavors by enormously advancing the capacity for synthetic research. The investigation of fundamental information integration challenges will thus contribute substantially to a shared infrastructure of science and will enable crucial transdisciplinary research concerning complex systems.Participation in this research by the computer science graduate students will prepare them to function effectively in multidisciplinary teams and enhance their appreciation of the associated challenges and opportunities. Use of UDI as a testbed will enable these students to experiment in scientific information management, thereby increasing their awareness of data integration and science-informatics issues. UDI We expect two graduate courses to leverage the data sets as well as the project software as an educational platform. Arizona State University also recruits top quality undergraduates through a nationally recognized residential Honors College and the Minority Access to Research Careers program and the project will involve undergraduate honors students to participate in the project. UDI will also serve as a testbed for undergraduate students through Capstone Projects.