A. Motivation Scenario

Currently, efforts to develop ontologies that generate recommendations for retrieving and publishing educational resources are being made. The work carried out in this regard has primarily focused on representing essential elements within specific domains. Below, we describe some notable works in this field, which were found through an exhaustive search in main research databases such as SCOPUS and IEEE:

1) Representation of Datasets. Ontologies are developed to represent and discover datasets that describe resources, e.g., Onto4AIR2 ^[7] illustrates that this straightforward ontology employs vocabularies like FOAF (Friend of a Friend) for metadata definition[8]; SKOS (Simple Knowledge Organization System) for the representation of controlled vocabularies, taxonomies, and thesauri ^[8]; schema.org provides a collection of vocabularies used to embed metadata in search engine-friendly web pages ^[8]; Onto4AIR2 facilitates the construction and management of machine-understandable datasets by allowing the addition of new terms, object properties, rules, or constraints. This approach improves the quality, reliability, and reusability of open repository data. In contrast, ^[9] focuses on the semantic representation of resources within DAJEE (Dataset of Joint Educational Entities) by applying the linked data principles and interconnecting many resources in the web of data. The DBpedia ontology is instrumental and establishes connections to further enrich the resources in DAJEE by providing categories and links considering the user's input. Daga ^[10], first presents a mapping for IEEE LOM metadata for linked data, and then an implementation of this study.

2) Resource Recommendation. These ontologies are primarily designed to recommend one or several educational resources relevant to a given search. For instance, an ontology centered on educational standards is developed, thus enabling teachers to simplify their lesson development process. It also serves as a support tool for students by establishing links to pertinent educational resources ^[11]. Moreover, these ontologies play a pivotal role in delivering personalized recommendations. In ^[12], the EduCOR educational ontology establishes a foundation for representing online learning resources within customized learning systems. This ontology empowers repositories of educational resources to offer tailored path recommendations aligned with the user's learning objectives, academic inclinations, and psychological considerations. Another notable effort is presented in ^[13], where the OntoSIDES ontology can track student progress. The outcome of this ontology is a taxonomy comprising 52 classes, encompassing seven main categories that define various potential student actions, types of pedagogical resources, a selection of medical references, French citations, milestones recorded by students as steps toward obtaining a diploma, and a classification of individuals involved in medical studies.

3) Use of Ontology Properties. The following works use pre-existing and tested ontologies, often combining various ontologies to index and upload sets of documents onto the web and leverage semantic links between resources and their underlying concepts. In ^[14], a collection of pedagogical resources is indexed, using links to facilitate inferences regarding relevant resources. The authors propose three ontology domains: information retrieval, e-learning, and knowledge engineering. For instance, they systematically organize the e-learning domain to improve information retrieval precision. Then, they harness semantic links between resources using extended inference rules. In ^[15], the DBpedia ontology's resources and categories are employed to identify related videos through a similarity function. Similarly, ^[16] uses the PAV (Provenance, Authoring, and Versioning) ontology, which offers provenance properties for resources, including essential metadata for their description. Likewise, ^[17] presents an approach for in-depth extraction of learning objects (LO) that yield pertinent outcomes from educational search engines. This approach relies on semi-automatic metadata detection and the resource relationships established by the proposed OntoAlgO ^[17] ontology to generate semantic indexing.

In the reviewed papers, we examine various proposals that use ontologies to improve the retrieval of specific types of educational resources on the web. However, these works often lack comprehensive insights into the essential metadata required for publishing and enhancing the retrieval process of educational resources. Furthermore, they may not delve deeply into generating tailored recommendations based on the proposed solutions.

B. Protocol to Construct the Proposed Ontology

A detailed review of methodological approaches to build ontologies such as Methontology ^[18], REFSENO (Representation Formalism for Software Engineering Ontologies) ^[19], Tove ^[20], Enterprise Model Approach ^[20], KBSI IDEF5 ^[20], and Noy and McGuinness ^[6] was carried out. The latter was selected for the construction of the ontology proposed in this article due to its versatility and results from previous experiences, Figure 1 summarizes it.

Fig. 1 Steps of the Noy and McGuinness Methodology. Adapted from ^[6]. 

The following subsections present the definition of the ontology and its implementation using the RDF markup language, which is the most important due to the simple form of its model for describing resources. This model is essential to store information efficiently, process it automatically, and exchange data between applications in triple form, each made up of a subject, a predicate, and an object. A set of such triples is called an RDF graph ^[21], which is represented in the Protégé tool.

The Protégé tool allows knowledge to be formally represented in the form of logic-based ontologies and with precisely defined semantics under the World Web (W3C), the standard language for web ontologies. Protégé has experienced a growing demand from users, both in industry and academia ^[22].

C. Purpose

The overarching goal of the ontology is to provide users with a guide or a set of steps for publishing and retrieving educational resources on the web. Doing so gives individuals interested in these processes an understanding of the required concepts and relationships, irrespective of their chosen approach. This ontology serves the following perspectives:

D. Competency Questions

Competency questions are formulated in natural language to ascertain the ontology's scope. These questions, along with their corresponding answers, helped extract the main concepts, their relationships, and properties within the ontology. It's important to note that no existing ontologies are aligned with the defined purpose; therefore, they have not been considered in this design. The proposed ontology aims to address the questions presented in Table 1.

E. Definition of Concepts, Properties, and Relationships

The key terms within the ontology resulting from the literature review, metadata of interest for the publication, and retrieval of educational resources are Publication, Owner, User, Publisher, Person, Educational resource, Retrieval, Resource type, Type of Site, Recommendation, and Metadata. Table 2 showcases the classes and their hierarchical arrangement, established through a combined strategy of generalization (top-down) and specialization (bottom-up) to identify the main concepts.

Fig. 2 Definition of class hierarchy. 

Table 3 describes the relationships that provide complementary information between the concepts.

Figure 3 defines the classes' properties along with their corresponding value types, illustrating the overall design of the proposed ontology.

Fig. 3 General design of the ontology. 

F. Implementation

Creating the Class Hierarchy.Figure 4 (a) illustrates the classes formulated under a larger entity, denoted by default as OWL:Thing in the Protégé tool. This hierarchy is established by considering the level of each class and its subclasses.

Definition of Data Properties.Figure 4Fig. (b) and (c) provide the list of properties, along with an illustrative domain and range example.

Definition of relationships.Figure 5 (a) and (b) depict the ontology relationships list, offering an example of one relationship's description. Each relationship between ontology classes is established while considering the associated cardinality. For instance, the cardinality between an educational resource and a user is many-to-many (the relationship must be completed separately), denoted as isSearchedBy and searchFor in Protégé.

Fig. 4 Implementation of the ontology classes in Protégé. 

Creation of Individuals.Figure 5 (c) showcases the generated individuals alongside their respective properties for a specific context. This step ensures that the ontology effectively addresses the competency questions outlined in Table 2.

Fig. 5 Implementation of relationships and visualization of individuals of the Ontology in Protégé. 

Figure 6 presents the Ontology implementation in Protégé as a graph, providing a visual depiction of the classes, their relationships, and hierarchies.

Fig. 6 Ontology knowledge graph.