Ontology

Ontology

The IOA-ORM has a formal ontology representation in three formats:

Format	File	Purpose
JSON-LD	ontology/context.jsonld	Machine-readable context for JSON-LD documents
Turtle (TTL)	ontology/exam-runtime-ir.ttl	Human-readable RDF serialization
TypeScript	spec/schema	Source of truth for runtime implementations

Namespace

@prefix exam: <https://exam-runtime-ir.org/ontology#> .

Ontology Statistics

Metric	Count
Classes	25
Properties	45
Individuals	8 (PromptingLevel instances)
PROV-O alignments	3 (Entity, Activity)

How the Three Layers Connect

TypeScript Interfaces          JSON-LD Context            OWL/Turtle
(running code)                 (linked data)              (semantic web)
┌──────────────┐              ┌──────────────┐           ┌──────────────┐
│ interface    │──validate──▶ │ @context     │──export──▶ │ rdfs:Class   │
│ ExamRuntime  │              │ exam:Exam    │           │ owl:Class    │
│ Package {…}  │              │ RuntimePkg   │           │ rdfs:prop    │
└──────────────┘              └──────────────┘           └──────────────┘
       │                            │                          │
       ▼                            ▼                          ▼
  Runtime impl.              JSON-LD documents           Reasoning,
  Validation                 External linking            SPARQL queries

Key Design Decisions

1. JSON-LD as bridge format: JSON-LD is both human-readable and machine-readable. It can be used directly in JSON documents as a context, and it maps cleanly to RDF/OWL.

2. PROV-O alignment: EvidenceSignal is a subclass of prov:Entity (provenance). RuntimeEvent is a subclass of prov:Activity. This enables standard provenance queries.

3. Joughin’s dimensions as first-class: The AssessmentProfile class encodes all 6 dimensions from Joughin (1998) as properties, making them queryable via SPARQL.

4. PromptingLevel as individuals: The 8 prompting levels (Pearce & Chiavaroli 2020) are OWL individuals, not just string values. This enables reasoning about which levels are permissible.

Using the JSON-LD Context

To make any JSON document semantically linked, add the context:

{
  "@context": "https://raw.githubusercontent.com/Stlouislee/ExamRumtimeIR-DSR/main/ontology/context.jsonld",
  "@type": "ExamRuntimePackage",
  "examId": "cs301-midterm-2026",
  "version": "1.0.0",
  "metadata": {
    "title": "CS301 Operating Systems Oral Exam",
    "assessmentPurpose": "summative"
  }
}

SPARQL Query Examples

Find all required evidence targets:

PREFIX exam: <https://exam-runtime-ir.org/ontology#>
SELECT ?target ?description WHERE {
  ?target a exam:EvidenceTarget ;
          exam:isRequired true ;
          exam:description ?description .
}

Find all evidence signals with high confidence:

PREFIX exam: <https://exam-runtime-ir.org/ontology#>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>
SELECT ?signal ?target ?confidence WHERE {
  ?signal a exam:EvidenceSignal ;
          exam:confidence ?confidence ;
          exam:targetId ?target .
  FILTER (?confidence > 0.8)
}

Find all events for a session:

PREFIX exam: <https://exam-runtime-ir.org/ontology#>
SELECT ?event ?type ?timestamp WHERE {
  ?event a exam:RuntimeEvent ;
         exam:eventType ?type ;
         exam:timestamp ?timestamp .
  ?session a exam:RuntimeState ;
           exam:sessionId "session-123" .
}

Tools

Tool	URL	Description
Ontology Graph	ontology-viz.html	Interactive force-directed graph of all classes, properties, and relations
SPARQL Query	sparql.html	Client-side SPARQL query engine with example queries
TS → OWL Generator	ontology/scripts/ts-to-owl.mjs	Generate OWL/Turtle from TypeScript schema
SHACL Shapes	ontology/shapes.ttl	Validation constraints for instance data

Future: SHACL Validation

When the ontology matures, we will add SHACL (Shapes Constraint Language) shapes to validate exam packages against the ontology:

exam:ExamRuntimePackageShape a sh:NodeShape ;
    sh:targetClass exam:ExamRuntimePackage ;
    sh:property [
        sh:path exam:examId ;
        sh:minCount 1 ;
        sh:datatype xsd:string ;
    ] ;
    sh:property [
        sh:path exam:version ;
        sh:minCount 1 ;
        sh:pattern "^\\d+\\.\\d+\\.\\d+$" ;
    ] .

This would enable runtime validation: “Does this JSON document conform to the IOA-ORM ontology?”