Research Knowledge Graph
OpenGraph converts research documents into a connected graph of papers, authors, institutions, and topics — so an agent can answer questions by walking the relationships, not just reading paragraphs.
Research is a graph pretending to be a pile of PDFs.
A single research question spans papers, authors, institutions, citations, and topics — but that structure lives implicitly across a folder of documents.
Vector search returns the paragraphs most similar to a query. It does not know that Paper B cites Paper A, that both authors work on adjacent topics, or which institutions are behind the work.
Extract entities. Connect relationships. Query the graph.
Parse research documents into typed nodes — papers, authors, institutions, topics.
Link entities with typed edges: authored, affiliated with, discusses, cites, related to.
Walk the graph to answer questions vector search can't — cited by, co-authored, related topics.
A slice of a research corpus.
A tiny sample of the kind of graph OpenGraph produces. Drag nodes, zoom, or select an entity to inspect its relationships.
Drag nodes · pinch or use +/− to zoom · tap a node to inspect it