Collaborative Research: Framework: Software: HDR: Building the Twenty-First Century Citizen Science Framework to Enable Scientific Discovery Across Disciplines

合作研究：框架：软件：HDR：构建二十一世纪公民科学框架以实现跨学科的科学发现

• 批准号: 1835632
• 负责人: Subhashini Sivagnanam
• 金额: $ 6.35万
• 依托单位: University of California-San Diego
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1835632&HistoricalAwards=false
• 关键词: Collaborative; Research; Framework; Software; HDR

A team of experts from five institutions (University of Minnesota, Adler Planetarium, University of Wyoming, Colorado State University, and UC San Diego) links field-based and online analysis capabilities to support citizen science, focusing on three research areas (cell biology, ecology, and astronomy). The project builds on Zooniverse and CitSci.org, leverages the NSF Science Gateways Community Institute, and enhances the quality of citizen science and the experience of its participants.This project creates an integrated Citizen Science Cyberinfrastructure (CSCI) framework that expands the capacity of research communities across several disciplines to use citizen science as a suitable and sustainable research methodology. CSCI produces three improvements to the infrastructure for citizen science already provided by Zooniverse and CitSci.org: - Combining Modes - connecting the process of data collection and analysis; - Smart Assignment - improving the assignment of tasks during analysis; and - New Data Models - exploring the Data-as-Subject model. By treating time series data as data, this model removes the need to create images for classification and facilitates more complex workflows. These improvements are motivated and investigated through three distinct scientific cases: - Biomedicine (3D Morphology of Cell Nucleus). Currently, Zooniverse 'Etch-a-Cell' volunteers provide annotations of cellular components in images from high-resolution microscopy, where a single cell provides a stack containing thousands of sliced images. The Smart Task Assignment capability incorporates this information, so volunteers are not shown each image in a stack where machines or other volunteers have already evaluated some subset of data. - Ecology (Identifying Individual Animals). When monitoring wide-ranging wildlife populations, identification of individual animals is needed for robust estimates of population sizes and trends. This use case combines field collection and data analysis with deep learning to improve results. - Astronomy (Characterizing Lightcurves). Astronomical time series data reveal a variety of behaviors, such as stellar flares or planetary transits. The existing Zooniverse data model requires classification of individual images before aggregation of results and transformation back to refer to the original data. By using the Data-as-Subject model and the Smart Task Assignment capability, volunteers will be able to scan through the entire time series in a machine-aided manner to determine specific light curve characteristics.The team explores the use of recurrent neural networks (RNNs) to determine automated learning architectures best suited to the projects. Of particular interest is how the degree to which neighboring subjects are coupled affects performance. The integration of existing tools, which is based on application programming interfaces (APIs), also facilitates further tool integration. The effort creates a citizen science framework that directly advances knowledge for three science use cases in biomedicine, ecology, and astronomy, and combines field-collected data with data analysis. This has the ability to solve key problems in the individual applications, as well as benefiting the research of the dozens of projects on the Zooniverse platform. It provides benefits to researchers using citizen scientists, and to the nearly 1.6 million citizen scientists themselves.This award by the Office of Advanced Cyberinfrastructure is jointly supported by the Division of Research on Learning in Formal and Informal Settings, within the NSF Directorate for Education and Human Resources.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.

来自五个机构（明尼苏达大学，阿德勒大学，怀俄明大学，科罗拉多州立大学和加州大学圣地亚哥大学）的专家团队链接基于现场的和在线分析能力，以支持公民科学，重点关注三个研究领域（细胞生物学，生态学和天文学）。 该项目建立在Zooniverse和Citsci.org上，利用NSF Science Gateways社区研究所，并提高了公民科学及其参与者的经验的质量。该项目创建了一个综合的公民科学Cyber​​infrasture（CSCI）框架，从而扩大了多个学科的研究社区的能力，以使用公民科学的研究来扩展公民科学的研究和可持续研究方法。 CSCI对Zooniverse和Citsci.org已经提供的公民科学基础设施产生了三个改进： - 组合模式 - 连接数据收集和分析过程； - 明智的分配 - 改善分析过程中任务的分配；和 - 新的数据模型 - 探索数据对象模型。 通过将时间序列数据视为数据，该模型消除了创建分类图像并促进更复杂的工作流程的需求。 这些改进是通过三种不同的科学案例进行动机和研究的： - 生物医学（细胞核的3D形态）。 目前，Zooniverse“ Etch-A-Cell”志愿者提供了来自高分辨率显微镜图像中细胞成分的注释，其中单个细胞提供了一个包含数千张切片图像的堆栈。 智能任务分配功能包含了此信息，因此在堆栈中未显示志愿者，在该堆栈中，机器或其他志愿者已经评估了一些数据子集。 - 生态学（识别单个动物）。 在监测广泛的野生动植物种群时，需要对单个动物进行识别，以估计人口规模和趋势。 该用例将现场收集和数据分析与深度学习结合在一起，以改善结果。 - 天文学（表征光曲面）。 天文学时间序列数据揭示了各种行为，例如恒星耀斑或行星跨性别。 现有的Zooniverse数据模型需要在汇总结果和转换返回之前对单个图像进行分类以参考原始数据。 通过使用数据对象模型和智能任务分配能力，志愿者将能够以机器辅助的方式扫描整个时间序列，以确定特定的光曲线特征。该团队探讨了重复的神经网络（RNN）的使用，以确定最适合项目的自动化学习体系结构。 特别有趣的是，邻近受试者的耦合程度如何影响性能。基于应用程序编程接口（API）的现有工具的集成也有助于进一步的工具集成。 这项工作创建了一个公民科学框架，该框架直接推动了生物医学，生态学和天文学中三种科学用例的知识，并将现场收集的数据与数据分析相结合。这具有解决各个应用程序中的关键问题的能力，并有助于对Zooniverse平台上数十个项目的研究。它为使用公民科学家的研究人员和近160万公民科学家本人提供了好处。该奖项由高级Cyber​​infrasture办公室颁发的奖项由NSF教育和人力资源局内的正式和非正式环境中的学习研究部门共同支持，这一奖项反映了NSF的法定任务和审查局长的范围。