Source-Traceable Knowledge Graph Traversal

Multiple graph lenses expose one evidence graph while every relationship retains its source.

RI.KG.001·Maintained reference·Revised 2026-08-06

A source-traceable knowledge graph connects companies, products, people, deals, markets, and claims. Every edge retains the sentence or record that supports it. Different traversal views operate on one graph rather than separate datasets.

629 company entities
11 graph lenses
~40k sourced facts

Graph lenses: One universe with bounded relationship filters

Entry modeStarting objectTraversal purpose
UniverseAll entitiesObserve clusters across relationship types.
EntityOne company or productFollow ownership, competition, and evidence.
AudienceOne stakeholder classFind entities serving the same users.
MarketOne clinical or commercial areaCompare participants and claims.

Traceability: A graph edge is an evidence claim

entity A → relationship → entity B → source pointer → supporting sentence
  • Retain the source pointer on every edge.
  • Separate extracted claims from inferred clusters.
  • Allow every visible relationship to open its evidence.
  • Use lenses as filters, not duplicate graph stores.

Traversal modes: Four entrances to one graph

Provenance key: Source identity joins facts across corpora

The source tag is a join key, not decoration. It links the visible fact to its file and line range. Shared claims can connect companies. External identifiers can extend the trail into PubMed, FDA, and SEC evidence.

node → labeled edge → node → fact identifier → file and line range → source sentence