Canonical graph v0.2.1 Engineering validation passed · research validity remains provisional
← Back to the Observatory

A number you can inspect

What the 1.98M graph edges actually represent.

The Phase-1 pipeline created typed, source-linked connections from 42,748 acquired work records. Each edge is one connection in the data model—not a human-reviewed research finding.

This is the publication metadata graph behind Module 1. It is separate from the cross-layer project-evidence pilot.

Graph composition

Most links are citation references.

Select an edge type to see its exact data-model meaning and its interpretation limit.

Displayed denominator 1,982,711 links

Citation links account for 62.35% of all graph edges.

1,236,187+333,427+289,560+123,537=1,982,711

How one edge is formed

Four relation types, four different meanings.

The work record provides the context. The same author–institution connection on two works is intentionally kept as two evidence-bearing observations.

What passed

Engineering integrity

  • Database integrity returned ok.
  • Edge count matches the database and exported JSONL.
  • Hashes, schema and graph closure passed 20/20 checks.
  • All acquired works reconcile with the canonical graph.
What remains provisional

Research interpretation

  • AI relevance still needs blinded human validation.
  • Author identities are not externally resolved at scale.
  • Metadata links do not establish quality or impact.
  • No ranking, capability or causal claim follows from edge volume.

Provenance and reuse

A publishable definition, with the limits attached.

Observed 19 September 2026 from OpenAlex metadata with ROR-normalized institution identities.

Validation ledger

Canonical pipeline
v0.2.1
Offline tests
35 / 35 passed
Graph checks
20 / 20 passed
Database integrity
ok

Edge file SHA-256

d4c9e17b6636ed1557d8608c82c0fb3d7264e3ad6b8f716c0ca6087cc0dc1f8f
Reuse the exact definitionsFour rows with counts, relation semantics and claim boundaries.
Download CSV Download JSON

Status: Graph structure, hashes and reconciliation checks passed. Research validity remains provisional pending blinded human review and adjudication. Do not use this graph for institutional rankings or claims about capability, quality, impact or causality.