Measuring R&D knowledge diffusion through large databases

本研究通过开发一种新的数据来源，使科学、工程和创新的新分析成为可能，从而增加了对联邦研究资助对经济产出影响的理解。该项目将通过收集和连接三种类型的公共数据：联邦政府拨款和合同、美国专利和许可协议，改进对研发知识传播的衡量。该项目帮助联邦政府向政策制定者、利益相关者和公民传达公共研发投资的影响。通过了解导致联邦政府研究商业化许可的因素，联邦政府可以更具战略性地投资其资金，提高研发投资的生产率，加速创新，增加美国的经济效益。经济竞争力。此外，该项目直接实现了建立一个对公民负责任的联邦政府的目标。该项目在所有数据类型（包括专利、出版物和奖励）共享的可扩展通用数据模式上构建了一个数据平台。收集和组合这些数据将创建一个独特的、新颖的大数据集。美国专利的政府利益部分将用于连接专利和联邦政府投资，这可以在拨款数据库和联邦采购数据系统中找到。专利许可是不透明的，而且不是普遍可用的；然而，技术许可协议的报告向美国证券交易委员会提交，提供了丰富的技术开发公共数据集。该项目将使用机器学习来提取这些提及，建立一个许可协议数据库。具体来说，该数据库将与USPTO许可数据库中有限的可用信息相结合，并与联邦投资类型相关联。这种方法代表了一种通过跟踪多个来源和时间点的知识流来衡量研发影响的新方法。该项目将展示大规模收集许可数据并将数据与支持联邦政府投资联系起来的可行性。结果数据集将向公众提供。

This research increases understanding of the impact of federal research funding on economic outputs by developing a new data source that enables new analyses of science, engineering, and innovation. The project will improve the measurement of R&D knowledge diffusion by collecting and linking three types of public data: federal government grants and contracts, U.S. patents, and licensing agreements. This project assists the federal government in communicating the impact of public R&D investments to policy makers, stakeholders and citizens. By understanding the factors that lead to commercial licensing of federal government research, the federal government can invest its funds more strategically and increase the productivity of R&D investment, accelerate innovation and increase the United States? economic competitiveness. In addition, this project directly addresses the goal to create a federal government that is responsive and accountable to its citizens. The project builds a data platform on an extensible, common data schema shared by all data types, including patents, publications, and awards. Collecting and combining these data will create a unique, novel, big data set. The government interest section of a U.S. patent will be used to link patents and federal government investments, which can be found in grant databases and the federal procurement data system. Patent licensing is opaque and not generally available; however, reports of technology licensing agreements are filed with the U.S. Securities and Exchange Commission, providing a rich public dataset on technology development. This project will use machine learning to extract these mentions, building a licensing agreement database. Specifically, this database will then be combined with the limited information available in the USPTO's licensing database, and linked to types of federal investment. This approach represents a new way to measure R&D impact by tracing knowledge flows through multiple sources and points in time. The project will demonstrate the viability of collecting licensing data on a large scale and linking the data back to the supporting federal government investments. The resulting data set will be made publicly available.