Skip to content

Connection reconciliation review

Report date: 2026-06-01
Scope: Full cross-layer reconciliation pass over existing YAML, resource markdown, and layer reviews — no web search, no new entities.

Pipeline: python3 scripts/refresh_all.py (post-edit)
Baseline snapshot: full-refresh-report.md from 2026-05-31 pre-pass counts (grant–resource 8/12 graph; person–output 14/15 graph).

See also: graph-readiness-review.md · grants-layer-review.md · publications-quality-review.md · outputs-quality-review.md

1. Files inspected

Category Paths
People & person–resource data/people.yaml, data/person_resource_links.yaml
Grants data/grants.yaml, data/person_grant_links.yaml, data/grant_resource_links.yaml
Publications data/publications.yaml, data/person_publication_links.yaml, data/publication_resource_links.yaml
Outputs data/outputs.yaml, data/person_output_links.yaml, data/output_resource_links.yaml
Courses data/course_resource_links.yaml, data/resource_courses.yaml, resources/CUR/CUR-036.md, CUR-038.md
Equipment & resources data/resource_equipment.yaml, resources/**/*.md (non-course)
Access / verification data/resource_access.yaml, data/resource_verification.yaml
Partner context resources/EXT/EXT-007.md, EXT-005.md, EXT-011.md
Automation scripts/refresh_all.py (sync_grant_resource_links_from_grants), scripts/export_graph_tables.py
Reports docs/reports/full-refresh-report.md, layer quality reviews

Automated cross-checks: registry people / related_resources arrays vs link YAML; resource_courses related_resources vs course_resource_links; graph-included publication/output–resource vs person–resource coverage.

2. Entity counts before and after

Entity Before After Δ
Resources (total) 265 265 0
Resources (non-course) 78 78 0
Courses (CUR-*) 187 187 0
People 115 115 0
Grants 33 33 0
Publications 34 34 0
Outputs 18 18 0
Equipment resource records 51 51 0
Equipment line items 345 345 0

3. Edge counts before and after

Edge type Total (before → after) Graph-included (before → after) Excluded (before → after)
person–resource 209 → 209 174 → 174 35 → 35
person–grant 33 → 33 33 → 33 0 → 0
grant–resource 12 → 12 8 → 9 4 → 3
person–publication 39 → 39 38 → 38 1 → 1
publication–resource 6 → 6 4 → 4 2 → 2
person–output 15 → 16 14 → 15 1 → 1
output–resource 12 → 12 12 → 12 0 → 0
course–resource 2 → 2 — —
facility–facility 1 → 1 — —

Evidence levels (all edges): direct 187 → 188 · affiliation 59 · topical 8 · excluded 42 → 41.

0

people.yaml associated_resources (209 entries) already match person_resource_links.yaml — build_person_resource_links.py is in sync.

Cross-layer check: graph-included publication–resource and output–resource edges do not, by themselves, justify new person–resource links (co-authorship / output authorship ≠ facility use). Karver (PUB-013 → CSH-017/018) already has person–resource links to CSH-017; no second lab edge added from co-authorship alone.

0

All grants.yaml people entries have matching person_grant_links.yaml rows (33/33).

0 new rows — 1 upgraded (same 12 total links).

link_id Change Reason
GRL-007 GRA-029 → CDM-011: related_topic / include_in_graph: false → supports / include_in_graph: true data/NIH STRONG BRIDGE Facilities.pdf evidence in grants.yaml names MedIX Lab co-directed by PI Raicu and states IMP/MedIX relevance to proposal purpose — explicit facilities narrative, not EXT-011 topical seed alone.

sync_grant_resource_links (from grants.yaml) refreshed grant_resource_links.yaml on refresh_all.py.

0

All publications.yaml people entries have matching person_publication_links.yaml rows (39/39).

0

All publications.yaml related_resources entries have matching publication_resource_links.yaml rows (6/6). Graph-included remain 4 (PRL-PUB-001, 002, 005, 006).

1

link_id Person → output Role Graph
POL-016 PER-024 James Bort → OUT-009 Beyond the Loan collaborator true

Evidence: resources/EXT/EXT-007.md — Beyond the Loan collaboration text names James Bort among DePaul faculty contributors cited in SPARK/Aon public materials.

Registry sync: outputs.yaml OUT-009 people array updated to include PER-024 (metadata only; not a new output record).

0

All outputs.yaml related_resources entries have matching output_resource_links.yaml rows (12/12).

0

resource_courses.yaml lists related_resources / related_facilities on many courses, but only CSE 201 and CSE 301 (CUR-036, CUR-038) have explicit ISE program + Ramsay news + catalog documentation matching the bar in course_resource_links.yaml. No additional course keys met explicit facility-use text in this pass.

0

Equipment is modeled in data/resource_equipment.yaml (not as separate graph edges). Spot-check:

  • 51 non-external resources with equipment blocks · 345 line items
  • 12 partner/engagement EXT-* and UNI-012 have no equipment blocks (expected)
  • 15 resources with empty equipment blocks are programs/centers without itemized public lists (e.g., UNI-001, MUS-002) — not corrected without new named items in source pages

No mis-assigned equipment IDs found in sample audit.

Candidate Layers Reason
GRA-007 → MUS-001 grant–resource Viola-bow equipment grant abstract does not name School of Music studio or MUS-001
GRA-016 → CDM-014 grant–resource Federated Learning URC abstract (truncated CSV) does not name HPCC or CDM-014
PUB-010 → CSH-013 publication–resource ZITO paper + news cite APS (EXT-003); Byrne/Gonzalez lab synthesis context does not state McGowan lab instrumentation for this scattering work
PER-074 → CSH-018 person–resource Joint Karver–Southern PUB-013 does not prove Karver uses Southern lab space
GRA-028 → UNI-007 grant–resource Cross-center Steans thread — related_topic correctly excluded (GRL-006)
GRA-029 → UNI-005 grant–resource NIH PDF excerpt names MedIX/IMP only; AI Institute remains topical (GRL-008 excluded)
Person–resource from grant_pi person–resource 25× UNI-001 grant_pi links remain excluded — ORS hub, not facility use
POL Wagner → OUT-008 person–output Plant SPARK page still does not name Wagner — POL-008 stays excluded
CDM-012 / VAD for Schrank games output–resource Faculty bio does not tie Dumpy/Pedandeck to VAD Lab facility text
Candidate Notes
GRA-028 → CDM-011 (GRL-005) Included in graph at medium confidence via EXT-005 VIDA collaboration text — acceptable but program-level; ORS facilities excerpt would confirm
GRA-027 → CDM-011 (GRL-003) Stays excluded — low confidence VIDA page co-location only
PRL-PUB-003, PRL-PUB-004 Stay excluded — STEM Center profile / Grice method-list inference
OUT-014–016 person links IRL collective student projects — intentional registry-only people arrays
Course ↔ CDM-004 / CDM-015 Many catalog courses mention cyber-physical or fabrication keywords; need catalog or program page quotes naming a specific facility per course

0 downgrades in this pass.

1 upgrade (grant–resource GRL-007) — see §6.

Existing exclusions unchanged: POL-008, PPL-033, PRL-PUB-003/004, grant_pi → UNI-001, GRL-003/006/008.

16. False-hub risks checked

Resource Post-pass degree (≥5) Assessment
CDM-011 5 (new) Now hub-listed: +1 graph grant–resource (GRA-029 MedIX facilities PDF). Justified; monitor before adding topical MedIX edges
UNI-003 6 Unchanged — three explicit SPARK output–resource edges
CDM-016 14 Unchanged — five Bluelight output–resource edges
CDM-001 9 Unchanged — three IRL output–resource edges
UNI-001 — No grant–resource edges; 25 grant_pi person–resource remain graph-excluded

No new UNI-001 grant–resource edges. No person–person collaboration edges invented.

17. Remaining warnings

From post-pass full-refresh-report.md:

Warning Disposition
CDM-002, CSH-008 Valid/intentional — no named contacts for graph person–resource
OUT-014, OUT-015, OUT-016 Valid/intentional — IRL team artifacts without named persons

Validation errors: 0

  1. ORS pilot import — Use internal-data-wishlist.md to add grant related_resources from proposal facilities sections (not CSV abstracts alone).
  2. Re-run reconciliation after any ORS drop — scriptable checks: registry arrays vs link YAML (this pass’s method).
  3. Course links — Add course_resource_links only when catalog or official program pages quote a facility ID (CSE 201/301 pattern).
  4. Neo4j/RAGMap — Import only include_in_graph: true edges; load evidence_level and permission_level when ORS fields exist; keep grant_pi and topical edges in a separate audit layer.
  5. Hub monitor — CDM-011 crossed degree 5 solely from GRA-029; do not add further MedIX edges without facilities text.

Strongest newly reinforced connections

Person → Grant → Resource (upgraded)

Person Grant Resource Evidence chain
PER-015 Daniela Raicu (PI) GRA-029 NIH STRONG BRIDGE CDM-011 VIDA/MedIX/IMP NIH BRIDGE Facilities PDF names MedIX Lab co-directed by PI; grant–resource GRL-007 now graph-included
Person Output Resource Notes
PER-023 Wagner OUT-009 Beyond the Loan UNI-003 SPARK Already POL-009 + ORL-007
PER-024 Bort OUT-009 Beyond the Loan (via UNI-003 / EXT-007) POL-016 added; Bort named on EXT-007 for Beyond the Loan; output–resource remains Wagner/SPARK press evidence

Person → Publication → Resource (unchanged; audit confirmed)

Person Publication Resource Graph
PER-074 Karver PUB-013 caspase assay CSH-017 PRL-PUB-005
PER-111 Southern PUB-013 caspase assay CSH-018 PRL-PUB-006
PER-103 Gonzalez PUB-010 ZITO EXT-003 APS PRL-PUB-002

Course → Resource (unchanged; already explicit)

Course Resource Source
CSE 201 (CUR-036) CDM-001 IRL ISE program page + Ramsay news + catalog
CSE 301 (CUR-038) CDM-001 IRL Same documented Forge sequence

Grant → Resource → Equipment (illustrative, not new edges)

Grant Resource Equipment context
GRA-030 Southern URC CSH-018 Southern lab fluorescence methods — equipment via resource_equipment.yaml on CSH-001/003, not separate grant–equipment edges
GRA-029 (upgraded) CDM-011 MedIX/IMP computing — equipment on CDM-011 resource page if present

Summary

Action Count
New link rows added 1 (POL-016)
Grant–resource upgraded (graph inclusion) 1 (GRL-007)
New person–resource / person–grant / pub / output–resource / course / equipment 0
Rejected weak candidates 9+ patterns
Entity count change 0

This pass prioritized graph strength over link volume. The registry was already well-aligned across layers; the main gap was latent explicit evidence (NIH facilities PDF for GRA-029→CDM-011; EXT-007 text for Bort→Beyond the Loan).

Regenerate: python3 scripts/refresh_all.py