Enterprise Knowledge Graph

Purpose

Query a small synthetic graph that keeps the source of each fact: how are two entities connected, and what is downstream of a change?

Variables

• Ventities (nodes)
• Etyped relations (edges)
• π(e)source of edge e
• d(u,v)length of the shortest path between u and v

Assumptions

• The graph is inspired by a collections operation; entities and sources are synthetic labels.
• Path queries treat edges as undirected; impact queries follow their direction.
• Every edge has a source (an edge without one would not be stored).

Model

Data this would need in practice

• Real entities and relations, with identifiers.
• Evidence for each entity-resolution decision.
• The real source of each fact.

Interpretation

A short path means a direct dependency. The impact set tells you what must be reviewed if a node, such as the collection policy, changes. Each edge on a path comes with its source.

Limitations

• Entity-resolution errors would propagate into every query.
• A graph holds only what was modeled.
• This graph is small and hand-made.