Graph analysis use cases

Date: 2026-06-01
Graph source: data/graph_exports/*.csv (from refresh_all.py)
Defaults: Prefer include_in_graph=true and evidence_level in (direct, high-confidence affiliation); exclude UNI-001 from hub rankings unless explicitly studying ORS anchoring.

---

A. ORS / research development

A1. Grants that connect people to technical resources

Field	Detail
User question	Which active grants tie PIs to named facilities or instrumentation?
Graph entities	Grant, Person, Resource
Edge types	PI_ON, SUPPORTS_RESOURCE (grant–resource), optionally CONNECTED_TO_RESOURCE
Evidence required	edges_grant_resource.evidence_file, grant title/funder in nodes_grants; facilities narrative in resource Markdown
Query pattern	Person–PI_ON–Grant–SUPPORTS_RESOURCE–Resource where all edges include_in_graph=true
Possible output	Table: grant_id, PI names, resource_id, relationship, evidence_level
False inference risk	Medium — topical grant–resource links; verify evidence_level ≠ topical before proposal language

A2. Resources with public evidence but no grant evidence

Field	Detail
User question	Where do we document equipment or programs publicly but have no grant–resource link?
Graph entities	Resource, Grant
Edge types	Absence of SUPPORTS_RESOURCE; presence of equipment in YAML (phase 2: RESOURCE_HAS_EQUIPMENT)
Evidence required	resources/*.md, data/resource_equipment.yaml; internal grants CSV for candidates not yet in grants.yaml
Query pattern	Resources with equipment lines AND zero incoming SUPPORTS_RESOURCE with include_in_graph=true
Possible output	Gap list for ORS enrichment (align with internal-data-wishlist.md)
False inference risk	High if interpreted as “unfunded” — may mean grant narrative not yet harvested

A3. Central hubs excluding administrative hubs

Field	Detail
User question	Which facilities are most connected in the technical graph (not ORS)?
Graph entities	Resource
Edge types	CONNECTED_TO_RESOURCE, SUPPORTS_RESOURCE, DOCUMENTS_RESOURCE, OUTPUT_USES_RESOURCE
Evidence required	Edge counts only; spot-check top hub pages
Query pattern	Degree centrality on Resource WHERE resource_id <> 'UNI-001' and edges include_in_graph=true
Possible output	Ranked hub list (compare to full-refresh § hub table)
False inference risk	Medium — high degree may reflect many weak affiliations; filter evidence_level=direct

A4. Resources that should be enriched with ORS or purchasing data

Field	Detail
User question	Which high-traffic resources lack amount, PI, or equipment purchase metadata?
Graph entities	Resource, Grant, Person
Edge types	High CONNECTED_TO_RESOURCE degree; missing SUPPORTS_RESOURCE
Evidence required	Refresh warnings; grants.yaml amount fields; ORS wishlist fields
Query pattern	Hub resources NOT IN grant–resource subgraph
Possible output	Prioritized enrichment queue for ORS staff
False inference risk	Low if framed as documentation gap, not operational failure

---

B. Faculty seeking collaborators

B1. People who bridge CSH and CDM technical resources

Field	Detail
User question	Who has strong evidence of using both CSH and CDM facilities?
Graph entities	Person, Resource
Edge types	CONNECTED_TO_RESOURCE with evidence_level=direct (or director/lead roles)
Evidence required	Per-edge evidence_text; DePaul profile URLs from nodes_people
Query pattern	Person with ≥1 CDM-* and ≥1 CSH-* resource edges, include_in_graph=true, exclude affiliation unless vetted
Possible output	Shortlist with role types and resource names
False inference risk	High if affiliation-only — two weak CyberLabs-style edges ≠ collaboration

B2. People with strong cross-unit technical evidence

Field	Detail
User question	Who connects multiple ownership prefixes with director/lead/support edges?
Graph entities	Person, Resource
Edge types	CONNECTED_TO_RESOURCE where edge_type in (owns_or_directs, leads_or_uses, supports, collaborates)
Evidence required	evidence_file, confidence=high
Query pattern	Count distinct prefix from nodes_resources per person; threshold ≥2 prefixes
Possible output	Ranked bridge candidates with edge types
False inference risk	Medium — collaborates requires specific evidence text per intake rules

B3. Candidate collaborators for a new sensor/instrumentation proposal

Field	Detail
User question	Who already touches sensing, imaging, or instrumentation resources we might cite?
Graph entities	Person, Resource, Publication, Grant
Edge types	CONNECTED_TO_RESOURCE, PI_ON, AUTHORED, DOCUMENTS_RESOURCE
Evidence required	Equipment taxonomy keywords in resource YAML; publication relevance text
Query pattern	Subgraph around resources matching topic/equipment filter → neighboring people
Possible output	Contact list + suggested facilities to name in proposal
False inference risk	High without topic filter — topical edges must not drive ranking

---

C. Students seeking facilities or courses

C1. Resources likely relevant to a proposal abstract (student project)

Field	Detail
User question	I need imaging + fabrication for a capstone — what facilities are documented?
Graph entities	Resource, Course (phase 2), Equipment
Edge types	COURSE_USES_RESOURCE, RESOURCE_HAS_EQUIPMENT, CONNECTED_TO_RESOURCE
Evidence required	Equipment categories, course link YAML with URLs
Query pattern	Keyword match on equipment taxonomy + course–facility edges
Possible output	Facility list with access notes from resource Markdown
False inference risk	Medium — keyword match ≠ guaranteed access; cite access classification

C2. Resources with equipment but no named people

Field	Detail
User question	Who do I ask about this lab if the site lists equipment but no contact?
Graph entities	Resource, Person
Edge types	Absence of CONNECTED_TO_RESOURCE with include_in_graph=true
Evidence required	Refresh warnings (CDM-002, CSH-008); generic emails on resource pages
Query pattern	Resources in equipment index LEFT JOIN person edges WHERE count = 0
Possible output	“Contact unit admin / listed email” + ORS enrichment flag
False inference risk	Low for “no named graph contact”; high if claiming “no staff”

C3. Find resources that support student-built electronics projects

Field	Detail
User question	Where can students build hardware projects with documented tools?
Graph entities	Resource, Course
Edge types	COURSE_USES_RESOURCE, equipment associations
Evidence required	resource_equipment.yaml, curriculum links
Query pattern	Equipment category ∈ {electronics, fabrication, …} AND (course link OR makerspace resource type)
Possible output	Resource cards with URLs and access fields
False inference risk	Medium — course catalog ≠ all student access paths

---

D. Department chairs / deans

D1. Underdocumented resources that appear in courses but not grants

Field	Detail
User question	Which teaching facilities lack grant or publication evidence?
Graph entities	Resource, Course, Grant, Publication
Edge types	COURSE_USES_RESOURCE present; no SUPPORTS_RESOURCE / DOCUMENTS_RESOURCE
Evidence required	course_resource_links.yaml, grant layer completeness note
Query pattern	Course-linked resources anti-join grant/publication resource edges
Possible output	Dean briefing: “documented for teaching, weak research evidence”
False inference risk	High if called “low impact” — may be teaching-only by design

D2. Cross-unit bridges (strategic interdisciplinary ties)

Field	Detail
User question	Where do CDM and LAS/CSH share people or outputs?
Graph entities	Person, Resource, Output, Publication
Edge types	CONNECTED_TO_RESOURCE, AUTHORED, CREATED_OUTPUT, OUTPUT_USES_RESOURCE
Evidence required	Output/publication titles and resource links
Query pattern	Same as B1/B2 with dean-facing summary metrics
Possible output	Unit-pair matrix of edge counts by evidence_level
False inference risk	Medium — count affiliation separately from direct

D3. People with grants but no publications or outputs

Field	Detail
User question	Which PIs are in the grants layer but lack demonstration artifacts?
Graph entities	Person, Grant, Publication, Output
Edge types	PI_ON without AUTHORED / CREATED_OUTPUT
Evidence required	Grant CSV harvest scope (seed layer incomplete)
Query pattern	Person–PI_ON–Grant OPTIONAL MATCH pub/output → filter null
Possible output	Publication discovery queue (not “inactive researcher”)
False inference risk	High — publications layer not exhaustive

---

E. Makerspaces and facility managers

E1. Find equipment clusters by unit or resource

Field	Detail
User question	What equipment categories cluster at CDM-011 vs CSH-003?
Graph entities	Resource, Equipment (phase 2 node)
Edge types	RESOURCE_HAS_EQUIPMENT
Evidence required	data/resource_equipment.yaml, taxonomy
Query pattern	Group by resource_id, category
Possible output	Equipment cluster chart for staffing/training
False inference risk	Low for inventory listing; medium for “duplicate capability” claims

E2. Publications or outputs that demonstrate a facility’s use

Field	Detail
User question	What evidence shows researchers actually use our space?
Graph entities	Resource, Publication, Output, Person
Edge types	DOCUMENTS_RESOURCE, OUTPUT_USES_RESOURCE, AUTHORED, CREATED_OUTPUT
Evidence required	evidence_text on publication/output–resource links (high bar)
Query pattern	Match resource_id → incoming doc edges with include_in_graph=true
Possible output	Evidence packet for annual report
False inference risk	Low when citing explicit edges; high if inferring from co-location only

E3. Weak-edge audit for a facility

Field	Detail
User question	Which person links to our resource are too weak for external claims?
Graph entities	Person, Resource
Edge types	CONNECTED_TO_RESOURCE where evidence_level ∈ (affiliation, topical, administrative)
Evidence required	Full edges_person_resource.csv rows for resource_id
Query pattern	Filter by resource_id and evidence_level
Possible output	Manager review list to upgrade or exclude edges
False inference risk	Low — descriptive audit

---

F. Proposal writers

F1. Find resources likely relevant to a proposal abstract

Field	Detail
User question	Which DePaul facilities match “wearable sensing + data science”?
Graph entities	Resource, Grant, Publication, Equipment
Edge types	Topic/equipment match + existing SUPPORTS_RESOURCE / DOCUMENTS_RESOURCE
Evidence required	Grant abstracts in YAML, equipment taxonomy, facility Markdown
Query pattern	Text/keyword retrieval → validate with graph neighborhood
Possible output	Facility appendix paragraph drafts with citations
False inference risk	High without retrieval citations — must use RAG + graph

F2. What evidence do we have that Cinespace is used for student or faculty outputs?

Field	Detail
User question	Can we claim Cinespace (CDM-016) in this proposal?
Graph entities	Resource CDM-016, Output, Person, Publication
Edge types	OUTPUT_USES_RESOURCE, CONNECTED_TO_RESOURCE, DOCUMENTS_RESOURCE
Evidence required	OUT-* edges to CDM-016; Project Bluelight pages; SCA handbook refs
Query pattern	MATCH (r:Resource {resource_id:'CDM-016'})<-[*]-(e) with evidence retrieval
Possible output	Bulleted claims each with URL/file citation
False inference risk	Medium — student film ≠ research instrumentation

F3. False-hub audit before citing “central” facilities

Field	Detail
User question	Is UNI-001 or SPARK the right anchor, or a specific lab?
Graph entities	Resource
Edge types	All types; compare UNI-001 vs CDM/CSH hubs
Evidence required	Graph export rules on UNI-001; graph-readiness-review.md
Query pattern	Hub ranking with UNI-001 excluded; flag grant_pi edges
Possible output	“Use CDM-011 / CSH-003 instead of ORS hub” guidance
False inference risk	High if UNI-001 included — administrative, not technical

---

Query implementation map

Use case IDs	Primary Cypher file section
A1, A3, F3	analysis_queries.cypher — grants, hubs, false-hub
B1, B2, D2	cross-unit bridges
C2, E1	equipment / no people
D3	grants without pubs/outputs
D1, C3	courses (phase 2 YAML import)
A2, A4, F1	gap / enrichment (combine graph + retrieval)
E2, F2	strong evidence doc edges

Templates: graph_queries/analysis_queries.cypher