PFI-TT: A Hybrid Scalable Data Management System Providing Deep Access to the Scientific Knowledge in Data Science

The broader impact of this project lies in its transformative potential for accessing scientific knowledge in the field of data science. The data management system, algorithms, and Knowledge Graphs (KGs) will improve the way new advances in data science are searched. By intelligently reformulating the search questions, users will experience a direct path to the most relevant results, significantly reducing the time spent on the usual trial-and-error searches, enhancing the productivity of data scientists and other consumers of scientific knowledge. The potential societal impact will be substantial, as the advantages introduced by the algorithms and systems could lead to broad and significant cost savings in research expenditures for government, companies, and universities. This approach challenges the current de-facto standard of keyword-search and Large Language Models (LLMs). The LLMs are sophisticated artificial intelligence models designed to understand and generate human-like text based on the input they receive. This research promises a more effective and resource-efficient method for accessing scientific knowledge. The project will provide valuable experiential learning for students, fostering skills in technology and entrepreneurship and empowering them to translate their research into successful, student-led, high-tech startups. The project envisions a novel solution to access knowledge in the scientific literature in the form of an intelligent and scalable hybrid data management system. The solution aims to provide users with a user-friendly interface for interacting with scientific concepts, accurately extracted and organized from a myriad of peer-reviewed scientific publications organized in a KG. This KG will surpass traditional approaches, maintained by experts by scale, depth, and freshness of knowledge, continuously learning from the latest peer-reviewed literature. The system's value proposition lies in its potential to achieve time savings of several orders of magnitude compared to current scientific search methods, while also delivering more comprehensive search results. The approach focuses on unsupervised extraction of the main search concepts and their structure from the state-of-the-art publications, organizing them into a hierarchy. This structured approach will simplify access to complex knowledge, enhancing the overall quality of search results and minimizing time spent on the search. The versatility of the approach allows for generalization to other domains, promising broad applicability and impact. 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.