[GRADE D -- Graph topology analysis of 107,480 RedactedEntity nodes (GOY-14 mining G25-G29)]
| Metric | Value |
|---|---|
| Total RedactedEntity nodes | 107,480 |
| With entity_value | 107,422 |
| With estimated_role | 45 |
| With constraints | 13 |
| With context_before | 45 |
| Resolved (LIKELY_IDENTITY) | 3 |
| Unresolved | 107,477 |
| Resolution rate | 0.003% |
| Source | Count |
|---|---|
| rhowardstone-redaction | 107,422 |
| pacer-courtlistener | 32 |
| fbi-vault | 13 |
| nydfs-db-order | 13 |
The overwhelming majority (99.95%) of RedactedEntity nodes come from rhowardstone's automated redaction detection across the DOJ document corpus. These 107,422 entities have entity_value populated but lack the rich metadata (estimated_role, constraints, context_before/after) that the 13 FBI and 13 NYDFS entities have.
The enhanced ghost score (G39) with visibility classification:
| Person | Ghost Score | Class | Redaction Prox. | Doc Mentions |
|---|---|---|---|---|
| Gaver (Amos Gaver/Gayer) | 31.0 (v2) | RESOLVED | 10 | 4 |
| AUSA Barnes | 40.0 | SHADOW | 4 | 1 |
| antonia barnes | 30.0 | OBSCURED | 3 | 1 |
| SA redacted | 20.0 | OBSCURED | 2 | 1 |
| Alex Juan Turner | 10.0 | OBSCURED | 1 | 1 |
| Jane Doe No 4 | 10.0 | OBSCURED | 1 | 1 |
| R. Alexander Acosta | 10.0 | OBSCURED | 1 | 1 |
| Rey Epstein | 10.0 | OBSCURED | 2 | 2 |
| Jeffrey Epstein | 0.006 | VISIBLE | 8 | 13,378 |
The inverse ghost score (G26) -- scoring RedactedEntities by how many named persons surround them -- confirms Ghost_0 as the most "surrounded" entity with 22 named persons nearby. Ghost_1 (12), Ghost_3 (10), and Ghost_4 (8) follow. The remaining entities in the corpus have 1 or fewer nearby named persons.
200 person pairs share RedactedEntity connections (capped), meaning the redaction network creates implicit connections between named persons who never directly interact but orbit the same concealed identities. 30 orphaned RedactedEntities have zero relationships -- completely disconnected from the graph.
WHAT THIS SHOWS AND DOES NOT SHOW: The 107,480 RedactedEntity nodes represent the largest unmined dataset in the project. The 0.003% resolution rate means the redaction network is almost entirely opaque. The 107,422 rhowardstone entities have entity_value (the redacted text snippets) but no analytical metadata. Future work could cross-reference entity_value against known person names to achieve bulk resolution. The FBI's 13 entities are the most analytically valuable because they have rich constraint metadata that enables identity narrowing.