SoD: Data and Meta-Data Integration Maintenance

这个研究项目的重点是开发一个数据集成和转换过程，支持可维护性，适应性和演变。数据集成系统是允许独立设计和开发的结构化数据的转换、集成和交换的软件系统。数据中通常微妙而复杂的相互依赖性可能会使此类系统的创建、维护和使用变得非常具有挑战性。PI与合作者Renee米勒（多伦多大学）一起提供了一个强大的工具库和机制，用于协调数据表示方式的语义差异，包括视图，映射和转换语言。主要重点是维护实现数据语义整合和共享所需的元数据。该项目开发了一个为进化而设计的集成和转换过程。这项研究将发展一种新的理论，元数据发现和适应的基础上，现代统计学习和数据集成过程的新理论，不仅支持自动化，但也可维护性，适应性和演变。这项研究的主要贡献将是开发一个设计过程，支持强大的数据共享，这是设计科学的一个重要方面。作为这项工作更广泛影响的一部分，预期成果将有助于通过制定模式和映射发现和管理任务的基准来加强研究基础设施。研究结果和基准将在项目网站（http：//www.cs.umd.edu/biggetoor/sod）上公布，以便利向各种科学界传播知识和工具。在这个项目中开发的方法将是特别有价值的科学家谁经常需要收集，管理和整合不同的数据集。研究人员还计划与行业合作者合作，以了解和满足行业需求，接收反馈，并促进技术转让。

This research project focuses on developing a data integration and transformation process that supports maintainability, adaptability, and evolution. Data integration systems are software systems that permit the transformation, integration, and exchange of structured data that has been designed and developed independently. The often subtle and complex interdependencies within data can make the creation, maintenance, and use of such systems quite challenging. The PI, with the collaborator Renee Miller (University of Toronto) have available a robust arsenal of tools and mechanisms for reconciling semantic differences in how data is represented including views, mappings, and transformation languages. The major focus is on the maintenance of the metadata necessary to achieve semantic integration and sharing of data. This project develops an integration and transformation process that is designed for evolution. The research will develop a new theory of metadata discovery and adaptation based on modern statistical learning and a new theory of the data integration process, that supports not only automation, but also maintainability, adaptability and evolution. The major contribution of this research will be the development of a design process that supports robust data sharing, an crucial aspect in the Science of Design. As part of the broader impacts of this work, the expected results will contribute to an enhanced infrastructure for research by developing a benchmark for schema and mapping discovery and management tasks. The research results and the benchmark that will be accessible on the project Web site (http:/www.cs.umd.edu/~getoor/sod) to facilitate dissemination of knowledge and tools to a variety of scientific communities. The methods developed in this project will be of particular value to scientists who routinely need to gather, manage, and integrate diverse data sets. The researchers also plan to partner with industry collaborators in order to learn from and address industry needs, receive feedback, and facilitate technology transfer.