EduCOR Ontology Overview

• EduCOR Ontology is an educational and career-oriented formalism that unifies digital learning metadata, user profiles, and labour-market skills for personalized recommendations.
• It employs a modular OWL 2 DL-based structure and FAIR principles to construct interconnected educational knowledge graphs supporting curriculum-driven guidance.
• The ontology integrates multi-source resources and user parameters to enable adaptive learning paths, with empirical evidence suggesting improved academic outcomes.

EduCOR Ontology is an educational and career-oriented knowledge representation formalism designed to unify the metadata of digital learning resources, user profiles, and labour-market skill requirements for multi-source, personalised learning experiences. Its specification, grounded in OWL 2 DL and the FAIR principles, enables the construction of educational knowledge graphs and underpins recommendation systems for dynamic, goal-aligned learning paths. Originally defined in "EduCOR: An Educational and Career-Oriented Recommendation Ontology" (Ilkou et al., 2021), and further formalised in "An Ontology for Representing Curriculum and Learning Material" (Christou et al., 6 Jun 2025), EduCOR constitutes both a domain-level schema and an integration framework for curriculum-driven repositories, online platforms, and third-party ontologies.

1. Scope, Motivation, and Rationale

EduCOR aims to bridge digital learning materials, granular learner profiles—including academic and psychological states—and the labour-market’s required skills. Its fundamental purpose is the formal, interoperable representation of educational metadata to support adaptive recommendation services. Target use cases include:

• Normalisation and integration of heterogeneous Open Educational Resource (OER) repositories using modular OWL schema.
• Representation of knowledge, preferences, learning goals, and psychological parameters per user.
• Explicit linkage from learning content through "Knowledge Topics" to marketable "Skills", which align with job-ontologies such as ESCO.
• Computational support for queries of the form: "Given a target job, a user’s current skill profile and psychological state, generate an optimal sequence of OERs to close skill gaps."

EduCOR’s ontology facilitates dense interlinking across multiple repositories, supports curriculum composition, competency-based navigation, and recommendation logic, achieving coverage and adaptability in gold-standard evaluations (Ilkou et al., 2021, Christou et al., 6 Jun 2025).

2. Modular Architecture and Core Conceptual Patterns

The ontology employs Modular Ontology Modeling (MOMo), splitting the schema into seven self-contained “patterns,” here summarised as Editor's term “EduCOR Modules”. Each module constitutes a logical group of OWL classes and properties:

Module	Principal Classes	Key Relations & Data Properties
Educational Resource	ec:EducationalResource, MultimediaData	coversTopic, hasResourceType, rating, difficulty
Knowledge Topic	ec:KnowledgeTopic, Theory, Methodology	hasTheory, employsMethod
Skill	ec:Skill	requiresTopic, mapsToOccupation (ESCO)
Test	ec:Test, Question, Answer	testKnowledgeTopic, scoreValue
Learning Path	ec:LearningPath, LearningOutcome, Goal	hasGoal, comprisesOutcome, outcomeForTopic
Recommendation	ec:Recommendation	recommendedResource, usesLearningPath
User Profile	ec:User, UserProfile, LearningPreference	hasProfile, educationalLevel, emotionalState

All modules are OWL 2 DL-compliant and annotated via OPLa. Concepts are defined using standard DL notation, e.g.

$$\textsf{LearningPath} \sqsubseteq \exists\,\textsf{hasGoal}.\textsf{LearningGoal}$$

$$\textsf{Skill} \sqsubseteq \exists\,\textsf{requiresTopic}.\textsf{KnowledgeTopic}$$

This modularity supports import and extension in external systems, such as the selective use of only Skill or UserProfile patterns.

3. Integrations and Cross-Domain Linkage

EduCOR is cross-linked to established ontologies and controlled vocabularies to maximise interoperability and semantics:

• IEEE LOM (educational alignments), LRMI (learningResourceType), CCSO (curriculumCourseSyllabus), schema.org types (LearningResource, Course, Assessment)
• SKOS for hierarchical KnowledgeTopic organization
• FOAF for ec:User alignment (ec:User ≡ foaf:Person)
• Dublin Core for metadata annotation
• ESCO for skill-to-occupation mappings

For topical navigation, SKOS concepts provide broader/narrower relationships for knowledge traversal, e.g.:

@prefix ec: <https://w3id.org/educor#> . @prefix skos: <http://www.w3.org/2004/02/skos/core#> . ec:KnowledgeTopic a owl:Class ; rdfs:subClassOf skos:Concept .

In curriculum graphs, modules reference media and events provide media, personas determine learning paths, and curriculum groups modules, supporting dense cross-linking. Recommended best practices include aligning edu-ont:Module to schema:Course, reuse of SKOS for topic hierarchies, and linking authors to FOAF.

4. Semantic User and Resource Modelling

EduCOR’s user model is built around ec:UserProfile, which aggregates:

• ec:AcademicParameter (e.g., educationalLevel, testScores)
• ec:PsychologicalParameter (emotionalState, motivation)
• ec:LearningPreference (contentType, sessionLength)
• ec:LearningGoal (target skill/job)

The resource model assigns ec:EducationalResource instances to one or more ec:KnowledgeTopic, annotating with accessibility, quality, and difficulty. Media forms are explicitly represented with MultimediaData. Skills are decomposed into KnowledgeTopics and mapped to occupations via job ontologies.

The resultant knowledge graph enables traversals such as:

Job → Skills → Topics → OERs

which operationalise both coverage and granularity in learning experiences.

5. Querying, Validation, and Evaluation Methods

EduCOR supports retrieval and reasoning via SPARQL queries over an RDF triple store. Representative queries include:

• Retrieval of all OERs for a given learning goal, sorted by difficulty:

SELECT ?oer ?title ?difficulty WHERE { ?goal a ec:LearningGoal ; rdfs:label "Data Scientist" . ?path a ec:LearningPath ; ec:hasGoal ?goal ; ec:comprisesOutcome ?out . ?out ec:outcomeForTopic ?topic . ?oer a ec:EducationalResource ; ec:coversTopic ?topic ; ec:difficulty ?difficulty ; schema:name ?title . } ORDER BY xsd:integer(?difficulty) LIMIT 20

• Personalised recommendation based on academic level and psychological state:

SELECT ?rec ?resource WHERE { ?user ec:hasProfile ?p . ?p ec:storedIn ?ap . ?ap ec:educationalLevel ?level . ?p ec:storedIn ?pp . ?pp ec:emotionalState ?emo . ?rec a ec:Recommendation ; ec:generatedForUser ?user ; ec:recommendedResource ?resource . FILTER ( (?level="beginner" && ?emo="tired") => EXISTS { ?resource ec:difficulty "easy" } ) } LIMIT 10

Evaluation proceeds via two approaches:

• Gold-Standard Schemata Mapping: Coverage measured by recall:

$$\mathrm{Recall} = \frac{|C_\text{EduCOR} \cap C_\text{Gold}|}{|C_\text{Gold}|}$$

Scores reported: 0.833 (OER-Commons), 0.857 (SkillsCommons), 0.875 (Merlot).

Full recommender system evaluation would compute precision and $F_1$:

$$\mathrm{Precision} = \frac{|C_\text{EduCOR} \cap C_\text{Gold}|}{|C_\text{EduCOR}|}\qquad F_1 = 2 \frac{\mathrm{Precision} \times \mathrm{Recall}}{\mathrm{Precision} + \mathrm{Recall}}$$

• Task-Based Competency Evaluation: SPARQL-based answers to six competency questions on source adaptability, labour-market integration, and user-centric personalisation. Execution confirmed full competency coverage on a Fuseki triplestore (Ilkou et al., 2021, Christou et al., 6 Jun 2025).

6. Use Case: eDoer Platform Deployment and Empirical Findings

EduCOR was deployed in eDoer, a prototype hybrid recommender system for Data Science careers. Key features in eDoer include:

• Skill Pattern integration for job-skill mapping from live labour-market feeds.
• Knowledge Topic Pattern enabling topic decomposition for each skill ("SQL Joins", "Data Cleaning").
• Learning Path Pattern for sequencing topics based on prerequisite relations.
• Educational Resource Pattern for OER normalization and multi-repository indexing.
• Recommendation + User Profile Patterns for learner-adaptive hybrid recommendations (incorporating user preferences, test scores, emotional state).
• Test Pattern supporting adaptive quizzes, with results updating academic profile and subsequent recommendations.

Field deployment in a Business Analytics course at the University of Amsterdam showed that 24 students who followed EduCOR-generated career paths outperformed peers in final grades. This suggests that alignment between labour-market skills and user profiles via EduCOR may be beneficial for measurable learning outcomes.

7. Implementation Considerations and Extensions

Specification in both (Ilkou et al., 2021) and (Christou et al., 6 Jun 2025) provides materialized RDF/Turtle graphs, modular OPLa annotations, and external linkage strategies. The ontology can be populated via CSV-to-RDF pipelines robust to missing values. Extensions can incorporate additional controlled vocabularies, support for further curriculum design patterns, or alignment with other learning resource schemas. Formal definitions follow OWL 2 DL, supporting automated reasoning about curriculum structure and user-task alignment.

By consolidating meta-data, learning goals, and market skills through modular, cross-domain, and user-adaptive patterns, EduCOR delivers a robust foundation for curriculum-driven, recommendation-enabled educational knowledge graphs. Its empirical coverage and interoperability underpin its adoption in platform integration contexts requiring semantic precision, extensibility, and measurable learner impact.