This paper proposes an ontology to automatic classification of learning materials to the Software Engineering knowledge domain. The Software Engineering Body of Knowledge (SWEBOK) was used to define the hierarchical structure of the knowledge area. The Rational Unified Process (RUP) was used to add the axioms to represent the relationships between concepts and to enable the reasoning to SWEBOK knowledge areas. Two testing scenarios were designed and experiments were performed. The results show that the ontology is able to classify and locate learning materials from the Software Engineering area, according to the desired area, role, artifact or task.
In this context, ontologies play an important role because they can be applied to provide a common shared understanding of an information structure among individuals or organizations, as well as be used to enable the knowledge domain reuse and make explicit assumptions of a domain [Noy and McGuinness 2010].
Ontologies can describe a hierarchy of concepts related by subsumption relationships, in this case, a taxonomy-driven concept; or a structure, where the axioms are added in order to express relationships between concepts and to restrict their intentional interpretations [Guarino 1998]. Through ontologies, hierarchical structures of themes related to the learning materials can be defined using a common vocabulary to the knowledge area. Furthermore, it is possible to add reasoning to this structure in order to help the automatic classification of learning materials within the defined hierarchy. The automated classification is relevant when people do not hold enough knowledge to identify the theme related to the learning materials due to lack of common vocabulary of the knowledge area. Software engineers can be mentioned as an example.
In this context, this paper aims to propose an ontology to automatic classification of learning
materials related to the Software Engineering knowledge area. The ontology aims to facilitate the
search for learning materials within the given domain. The Software Engineering Body of
Knowledge (SWEBOK) [
The remainder sections of this paper are organized as follows: Section 2 presents the related work; Section 3 describes in details the proposed ontology; in Section 4 some experiments are discussed; Section 5 concludes the paper.
There are several papers proposing ontologies for the Software Engineering area. This section presents these researches and their approaches.
More specific approaches are established on the Software Engineering domain as well. The
Win-Win approach represents a model created to manage the necessary collaboration and negotiation
by the people involved in the software lifecycle stage [
There are several proposals for ontologies in the Software Engineering area, however, there is not an ontology to classify materials according to the Software Engineering knowledge area. The next section discusses the proposal of an ontology to help solving this problem.
This section presents an ontology composed by SWEBOK and RUP concepts to classify learning
materials in the Software Engineering knowledge area. The ontology was developed with the
ontology editor Protégé [
To define the knowledge’s hierarchical structure related to Software Engineering, the
SWEBOK´s definition knowledge area was used. The SWEBOK is a guide created under the
patronage of the Institute of Electrical and Electronics Engineers (IEEE) with the objective of
serving as reference to Software Engineering related subjects [
However, the definition of a hierarchical structure is not enough to allow the automatic classification of learning materials according to the defined structure. The SWEBOK does not present an approach to the definition of their knowledge areas using relationships among the concepts or explicit properties. For this reason, RUP was also used. RUP presents well-defined relationships among the main concepts, which are: Discipline, Artifact, Role and Task. Although RUP is a software development process, hence, not exactly focused on knowledge areas, the concept of disciplines can be related between some SWEBOK knowledge areas, as shown in Table 1. In this proposal, only the areas with total correspondence were mapped.
In the following subsections the details of the proposed ontology for RUP and the integration of this ontology with the ontology for the classification of learning materials according to the SWEBOK knowledge areas are presented.
OntoRUP was developed according to the Artifact, Role and Task concepts and their relationships with the Discipline concept. Through these four concepts and their relationships, classes and their properties were created. Table 2 presents the created classes and properties.
The general proposed hierarchy is presented in Figure 1. The RupElements class was created in order to group the derivative concept classes: Discipline, Artifact, Role and Task concepts.
The Discipline class was created to represent the nine disciplines that compose the RUP model. Through this class the other relationships are established and then the integration is done with the SWEBOK´s knowledge areas.
The Artifact class was created to represent the software artifacts that are used within the RUP process. The Artifact class is directly related to the Discipline class through the hasDomain property. According to this relationship, subclasses were created, that identify the artifacts related to each of the nine disciplines proposed in the RUP model. Figure 2 presents an example of the hasDomain property.
The Role class was created to represent the corresponding subclasses to the six groups of roles within the RUP, namely: Analysts, Developers, General Roles, Manager, Production Support and Testers. Furthermore, within the Role class, corresponding subclasses of the roles related to each of the nine disciplines were also created as shown in Figure 3. To establish the relationship between the Role and Discipline classes, it was used the property “modify” that relates the Role class to the Artifact’s subclasses. As the subclasses of Artifact are already related to the Discipline class, the relationship between the Role and Discipline classes is also completed.
The Task class was created to represent the tasks of the RUP model. The Task class has direct relationship with the Discipline class through the hasDomain property. Based on this relationship have been created subclasses to represent the tasks corresponding to each of the nine RUP disciplines specified in the model.
Once established the ontology structure for representation of RUP elements, it was defined the necessary elements to enable the classification of learning materials within the Software Engineering domain. The LearningMaterial class was created to represent the learning materials, and its subclasses were created based on the ten SWEBOK’s areas, as shown in Figure 4.
In order to define the ten areas of the SWEBOK using explicit and formal properties, the defined concepts of RUP ontology was used. It is possible to identify the related discipline through any of the Artifact, Role and Task concepts, and through mapping it is possible to know the SWEBOK´s knowledge area. Because of that, the isRecommendedTo property was created, as shown in Table 3, in order to be able to recommend a learning material related to any of the three concepts presented in RUP. Thus, when adding a learning material it is possible: to recommend the material for the use of a specific artifact, such as a Business Case; the execution of a specific task, such as Architectural Analysis; or the execution of a specific role, such as System Analyst.
Through the related recommendation it is possible to classify the material according to the SWEBOK’s knowledge areas. For instance, a learning material will be classified as belonging to the Test knowledge area, if it has the isRecommendedTo property related to, at least, one instance of the Artifact, Role or Task classes, linked to the Test discipline.
These possibilities of recommendations can help to obtain a more accurate classification of the learning material, especially when there is no formal knowledge regarding to which knowledge area the material belongs to.
The ontology was proposed to be applied in a self-learning environment where people share their knowledge related to the Software Engineering area by adding learning materials. The proposed ontology will help in the classification of learning materials, mainly because software engineers may not use a common vocabulary or may not have enough knowledge to classify correctly the material within the appropriate domain. Furthermore, the ontology will facilitate the recommendation of these learning materials.
Two scenarios were designed to verify the proposal’s viability. The scenario 1 was used to test the classification of learning materials and scenario 2 was used to test the recommendation of these materials. The simulations were created using the Protégé tool.
Scenario 1 – Learning Materials Classification
Instances of learning materials were added using the Protégé tool, as shown in Figure 5. Also recommendations were made through the isRecommendedTo property. Each recommendation was associated with instances of Artifact, Role or Task classes.
The values assigned to the isRecommendedTo class for each one of the learning materials are shown in Table 4.
The Pellet reasoned, version 1.5.2, was used to classify the learning materials. As shown in Figure 6, it is possible to verify that the ontology correctly classified the learning materials according to the defined concepts.
However, it is important to provide mechanisms to help software engineer to make their recommendations in order to avoid inconsistencies. For instance, the material identified as “material008”, was recommended to be used in the Business Case artifact. In this case, it should not be possible to recommend it for the System Analyst role, as this role has no relationship with this artifact. As a result, the material was classified in three knowledge areas, one of them due to artifact recommendation, and the other two due to recommendation by role. The ontology proposed can be used to help filter consistent recommendations among Artifact, Role and Task classes.
Scenario 2 was designed to present the possible recommendations of the learning materials once these materials will be available in a learning environment. According to the simulation described in scenario 1, after the learning materials were classified using the inference mechanisms, it is possible to search for these materials through the knowledge areas defined in SWEBOK. For instance, it is possible to retrieve all the learning materials related to the Software Requirement area. However, besides retrieving the materials by Software Engineering knowledge area, the ontology also allows to find all the materials according to recommendations, by Artifact, Role or Task.
SPARQL was used to simulate a preview of these possibilities. The SPARQL is a language to retrieve data from Web Ontology Language (OWL) files. Figure 7 presents a SPARQL query in order to retrieve learning materials recommended by Roles. In this case, the learning materials are retrieved through the Roles view; however, the queries can be executed by Artifacts and Tasks as well.
It is important to point out that new recommendations may be added to the learning materials according to their use. For example, a learning material that was added with the System Analyst role may also be recommended to the Elicit Stakeholder Requests task. So, the level of details for the recommendation is enhanced and the retrieval of material becomes more precise.
This paper presented an ontology to automatically classify learning materials related to the Software Engineering knowledge area, aiming to facilitate the search for these materials.
The ontology was defined using the main structure of ten SWEBOK knowledge areas and the concepts and relationships among Artifact, Task and Role elements from RUP model. RUP was used to define SWEBOK knowledge areas through axioms to enable the automatic classification of learning materials according to recommendations.
Some experiments were performed and it was possible to conclude that the ontology classifications were correctly, according to the Software Engineering knowledge areas. Furthermore, the ontology provides views of the learning materials under three aspects, recommendations by artifacts, tasks and role. This diversity can be another facilitator for retrieving the desired material.
The proposed ontology will be integrated to a self-learning environment, and experiments with Software Engineering students and professionals will be performed in order to evaluate the proposal.
Deridder, D. (2002). “A Concept-Oriented Approach to Support Software Maintenance and Reuse Activities”. In: 5th Joint Conference on Knowledge-Based Software Engineering (JCKBSE), Maribor, Slovenia.
Wongthongtham, P. (2006). “A methodology for multi-site distributed software development.” PhD Thesis, Curtin University of Technology.