CDI-Type II: Understanding the Universe through Scalable Navigation of a Galaxy of Annotations

不断增长的天文数据冲击带来了巨大的挑战。天文学缺乏一种易于使用和可扩展的方式，从数万到数十亿个单独事件和物体的数据集中收集和分发有关物体的专家信息。在接下来的十年里，典型的天文学家可以获得的信息量将在原始数据的大小和天体的数量上增长两个数量级。该项目追求三个研究方向，每个方向都有可能改变天文学家与大型数据集的接口方式：(1)可扩展的注释框架，使观测与特定实验、模型或其他观测相关联；(2)在存在不确定性的情况下支持自动推理的连续工作流制定系统；(3)交互式分析天文数据集的计算框架，允许构建和测试假设。项目计划包括设计和开发一个原型系统（AstroShelf），并使用两个及时的科学计划对其进行评估：(a)使用来自DEEP3和AEGIS调查的多波长数据，开发方法来整合来自不同数据集的图像和目录，使我们能够研究星系的人口统计数据在过去80亿年中是如何变化的。(b)利用泛星遥感调查发现的时变事件的特性，发展快速分类瞬变现象的技术，传播其特性，并纳入后续观测的反馈。AstroShelf将包括一个可公开访问的、灵活的公共数据集注释系统，这也将使它本身成为涉及普通公众注释的扩展努力。这个项目的重大影响是能够快速和广泛地分享信息和专家意见，关于每一个新的观察或事件，从根本上改变我们了解宇宙的能力；这种功能也可以应用于支持其他科学领域。这项研究的结果将在www.astroshelf.pitt.edu上公开。

The growing onslaught of astronomical data available presents a great challenge. Astronomy lacks an easy-to-use and scalable way to collect and distribute expert information about objects from datasets of tens of thousands to billions of individual events and objects. Over the next decade, the amount of information available to the typical astronomer will grow by two orders of magnitude both in raw data size and in the number of objects. This project pursues three research directions, each of which has the potential to transform how astronomers interface with large datasets: (1) a scalable annotation framework to enable linking of observations to specific experiments, models, or other observations; (2) a continuous workflow enactment system that would support automated reasoning in the presence of uncertainty; and (3) a computational framework for interactively analyzing astronomical datasets, allowing the construction and testing of hypotheses.Project plans include the design and development of a prototype system (AstroShelf) and its evaluation using two timely science programs: (a) Using multi-wavelength data from the DEEP3 and AEGIS surveys, develop methods to incorporate images and catalogs from disparate datasets, allowing us to study how the demographics of galaxies have changed over the last 8 billion years. (b) Using properties of time-variable events found by the Pan-STARRS survey, develop techniques for rapid classification of transient phenomena, dissemination of their properties, and incorporation of feedback from follow-up observations. AstroShelf will include a publicly accessible, flexible annotation system for public datasets, which will also lend itself to outreach efforts involving annotations by the general public. This project's significant impact is the ability to share information and expert opinions quickly and widely, about each new observation or event, fundamentally changing our ability to learn about the Universe; such functionality can also be applied in support of other scientific domains.Results of this research will be made publicly available at www.astroshelf.pitt.edu.