Science Genome: A Scholarly Graph Embedding Framework to Uncover the Fundamental Dynamics of Scientific Enterprise
Abstract
Research problem. Discovering the regularities, or laws, that govern the evolution of scientific enterprise and innovations holds great potential to efficiently nurture a successful national scientific enterprise, which is vital for the prosperity, well being, and defense of a nation. We identify the lack of unified quantitative frameworks as a key limitation of the current practices in science of science. Therefore, we propose to create a new quantitative framework, SCIENCE GENOME, that unifies the representation of all types of entities and encompasses coherent measurement and computation. Leveraging SCIENCE GENOME, we propose to discover the laws that govern the evolution of scientific enterprise and innovations. Proposed methods. We will develop and apply graph embedding methods such as Graph Convolutional Networks to obtain coherent vector space representations of scientific entities such as authors, papers, and concepts. Our approach is a major departure from prevailing approaches and has a potential to create a ground breaking platform that enables coherent quantitative inquiries that encompass multiple entity types across multi layered, multi scale, and heterogeneous scholarly networks. Furthermore, by aligning embeddings of temporal snapshots, we will model the temporal evolution of scientific enterprise as the movements and creations of points in the space, where physical analogies can be readily made and the conceptual hypotheses can be translated into quantitative inquiries. Anticipated outcomes. This project will create methods to embed entities in complex, heterogeneous scholarly graphs into a vector space as well as resulting representations of the entities (SCIENCE GENOME), which can become a fundamental basis for formulating quantitative inquiries on science of science. Successful completion of this project will then validate hypotheses on the evolution of fields and the dynamics of scientific discoveries formulated on the SCIENCE GENOME.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 14, 2022
- Source ID
- FA95501910391
Entities
People
- Yong Yeol Ahn
Organizations
- Air Force Office of Scientific Research
- Indiana University
- United States Air Force