Figure 3.1 shows overall system architecture. There are three modules in our learning sequences construction system. The Ontology-based Learning Sequences
Construction Module and the Meta-Knowledge Extraction Module are preprocessing
modules. And, the Example & Quiz Annotation Module would be used to integrate the preprocessing results into a recommended course scheme. The working principles of these modules will be covered in more details later in this chapter.
Figure 3.1: Overall system architecture
3.1 Learning Sequences Construction Preprocessing
In this thesis, we adopt both domain ontology and domain knowledge rules to construct learning sequences. In essence, ontology representation focuses on concept classes, attributes of the concept classes, and the relationships between the concept classes. On the other hand, the main components of rule representation are facts and actions. Thus, in practice, the integration of ontology and rules is not a straightforward job. Before integrating these two kinds of knowledge representations, it is supposed that there should be preprocessing processes to make the integration process smoother.
In essence, as mention in Section 2.1.1, we know that an ontology is an explicit specification of a conceptualization, and it could be used to model domain knowledge.
In addition, an ontology consists of concept classes and relations. Hence, we would like to transform the concept hierarchy and relationships between the concepts into some kind of learning sequence. Therefore, the approach to transform ontology into a course scheme (a set of learning sequences) is suitable. In this thesis, we propose an
Ontology-based Learning Sequences Construction Module to transform the domain
ontology into a basic course scheme. In essence, there are three main issues:(1) How to decide the transformation mapping between relationships of the domain ontology and learning sequences?
(2) When we confirm the transformations between relationships and learning sequences, the next problem is: how to decide the priority among these transformations?
(3) How to create a course scheme based on the definition of transformations and the priority ordering definition?
The working details in the Ontology-based Learning Sequences Construction Module will be described later in Chapter 4.
In general, rule representation is appropriate for the support of decision-making on network system management. In fact, many network services (e.g., IDS, anti-SPAM software, information filtering systems, etc.) adopt rules as the engine to perform access control jobs. For example, most firewall software systems are typical rule-based systems. As similar to other network management domain, rule representation is suitable for DNS domain as well. In these domains, adding rule-based knowledge is a very efficient way to enhance the content of course.
Usually, rule format was written as follows:
IF <Condition> THEN <Conclusion>;
In order to integrate rule-based knowledge into the course exactly and smoothly, we should get some information from this kind of representation, including:
z What are the key terms of the rule?
z Where is the most appropriate position of the domain course to add the rule-based knowledge?
Therefore, we propose to design the Meta-Knowledge Extraction Module to get the necessary information. In essence, this module has two intentions:
(1) Extract meta-knowledge from rules.
(2) Let the meta-knowledge be integrated with ontology knowledge smoothly.
The working details of this module will be covered later in Chapter 5.
3.2 Knowledge Integration
After the preprocessing of ontology knowledge and rule-based knowledge, we could get a basic course scheme and meta-knowledge of rules. Next, in order to offer learners more complete domain knowledge, we will integrate these two kinds of knowledge by adding the meta-knowledge into the basic course scheme. However, we would meet a problem: What form would we add the rule-based knowledge into the ontology knowledge with?
According to the pedagogy theory, we know that learning by examples or quizzes could increase the learning efficiency. In essence, the rules in DNS diagnosis are used to diagnose DNS problems and it is suitable to provide examples when DNS problems occurred. Furthermore, the facts and actions of DNS diagnosis rules are composed by DNS ontology elements. Thus, the rules could provide us the hits to provide appropriate examples at suitable place in the course scheme. In this thesis, we propose the Example & Quiz Annotation Module to annotate the rule-based knowledge (i.e., the meta-knowledge extracted from rules) into the ontology knowledge (i.e., the basic course scheme transformed from the domain ontology) with the form of examples and quizzes.
In practice, examples would usually be presented after learners have finished most of the related chapters (or sections). If we want to annotate the meta-knowledge of a rule as an example of the course, we have to know which chapter (or section) is the last one among those related chapters (or sections). Here we will use the learning sequences transformed by the Ontology-based Learning Sequences Construction Module to decide which one is the last studied related chapter (or section) in a rule.
In this thesis, we adopt the methodology proposed by Fischer (2001) to perform the quiz annotation. The main idea is to change the major conceptual terms with the same (or similar) relationships in the ontology to generate some simple, but meaningful quizzes. The quizzes include true-or-false, single choice, and even multiple choices. Both annotation methodologies will be described in Chapter 6.
3.3 Course Refinement
The course scheme generated by domain ontology and rules is a recommended one. In practice, teachers or domain experts could use some authoring tools to refine the course scheme for their own uses (i.e., to fit their own requirements). Because the course scheme follows the standard of SCORM 2004, the most popular standard for learning contents, there are many authoring tools to use. For example, Reusable eLearning Object Authoring & Delivery (RELOAD) provides a Metadata and
Content Packaging Editor, which can help users organize, aggregate and package
learning objects in standard IMS and SCORM content packages. Furthermore, in Su et al. (2005), the authors proposed an Object Oriented Course Modeling (OOCM) to construct the SCORM compliant course and supported a graphic OOCM authoring tool as shown in Figure 3.2.Moreover, based on our proposed architecture, system administrators can add additional information to make the course scheme more adaptive. For example, system administrators may add user profiles into the Ontology-based Learning
Sequences Construction Module to influence the definitions of transformations and
the priority ordering. In this way, the system could be used to generate moreindividualized course schemes.