Building an AI-Driven Research Knowledge Graph for Problem Discovery and Organization

Keywords: AI-driven research discovery, research knowledge graph, large language models, research problem extraction, semantic retrieval, citation-based relations, automated hypothesis generation, ranking mechanisms, autonomous research agents, human-in-the-loop validation, reproducibility, mechanistic interpretability

TL;DR: We propose an AI-driven system that extracts and organizes open research problems into a knowledge graph, enabling structured queries, ranking, and agentic extensions to accelerate scientific discovery.

Abstract: The rapid growth of scientific literature makes it increasingly difficult for researchers to identify open problems and track evolving opportunities. This paper proposes a vision for an AI-driven system that ingests research papers and transforms their content into structured, machine-navigable representations of open problems. By representing problem statements, assumptions, datasets, and constraints in a graph with semantic and citation-based relations, the system would enable novel queries and ranking mechanisms to surface high-value research opportunities. Importantly, this work is presented as a proposal and conceptual framework for future development rather than a description of a completed implementation. The contributions of this proposal lie in outlining the motivation, potential methodology, and expected impact of building such a research knowledge graph to support discovery, education, and collaboration across domains.

Submission Number: 173