Capítulo 2: Concepciones metodológicas de la investigación y resultados del trabajo de campo.
3. Análisis de los resultados de la encuesta:
Authoring content for adaptation has been around for as long as AH systems have existed. However, as we have seen in section 2.5, in the past, content often contained embedded adaptation. However, in order to be able to reuse both the content and the adaptive specification, a separation of concerns between content and adaptive specification is essential. This section introduces some methods for authoring content for AH, rooted in the separation of concern paradigm. First we describe the CAF format [46], which is based on the LAOS model and describes the content and content-related metadata, populating the Domain Model and Goal and Constraints Model layers, free from adaptive specification. Authoring the content, free from adaptation, is closely related to modelling the specific subject domain conceptually. Therefore this represents an ontology-based method of describing the domain content.
2.6.1 Common Adaptation Format (CAF)
The Common Adaptation Format [46] is a portable, system-independent XML format. It contains common elements related to the way adaptive content is represented in many AH systems. CAF was built to represent a part of the LAOS Adaptive Hypermedia model for authoring, and as such it is a very useful format for sharing and deploying authored AH courses. A CAF file describes a single AH course. The single course is represented by one
Goal Map from the LAOS framework and all Domain Maps (at least one) that are used in
the Goal Map. Each Domain Map consists of a hierarchy (tree) of domain concepts. Concepts have sub-concepts as well as domain attributes. The actual content is included in
40 these attributes. In addition to this, Domain Models allow the use of the relationlink element, which in a Domain Map indicating relatedness between two concepts. The Goal Map is also a hierarchy of lessons and sub-lessons. It contains semantic information for use in adaptation strategies in the form of weights and labels. Each of the lessons also contains a number of links, each linking to Goal Map a specific attribute of a concept in one of the Domain Maps.
The CAF format has one important limitation. The format is built under the assumption that the content is in fact a hierarchical tree. This is an important assumption, as hypermedia in general is a set of nodes and links, but not necessarily ordered as a hierarchical tree. Still, many AH systems assume a tree structure. This is due to the fact that many AH applications are applied to education, where a tree structure is used to represent content using similar ordering and structures to those in a text book. However, for the general case, more complex structures, such as graph-based structures, need to be considered.
Various systems have implemented the CAF format such as MOT [42], Content-e/LAOS [49] and AHA! [46]. The DTD definition of the CAF format is shown below. A CAF file consists of two parts, a domainmodel part containing a number of Domain Maps, and a goalmodel part containing the exported course. Below, the CAF DTD18 is shown.
<!ELEMENT CAF (domainmodel?, goalmodel?)>
<!ELEMENT domainmodel (concept+)>
<!ELEMENT concept (name, attribute*, concept*, relation*)> <!ELEMENT attribute (name, contents)>
<!ELEMENT name (#PCDATA)> <!ELEMENT contents (#PCDATA)>
<!ELEMENT relation (name, relationlink+)> <!ELEMENT relationlink (#PCDATA)>
<!ATTLIST relationlink weight CDATA ""
18
41 label CDATA ""
type CDATA "">
<!ELEMENT goalmodel (lesson)> <!ELEMENT link (#PCDATA)>
<!ATTLIST link weight CDATA "" label CDATA "">
<!ELEMENT lesson (link*, lesson*)>
<!ATTLIST lesson weight CDATA "" label CDATA "">
As the DTD shows, a CAF file consists of a Domain Model part and a goal model part. The Domain Model part consists of all concept maps (using the concept tag) needed for the goal model. The concept maps in the Domain Model consist of a name, a number of attributes, relations and sub concepts. Each attribute has a name and textual contents, be it a link, HTML or plain text. A relation has a name, and at least one relationlink to the related item, containing a weight, label and type.
The Goal Model itself, consists of exactly one lesson. Each lesson consists of a number of links to attributes from certain concepts in the domain maps, and a number of sub lessons. To exemplify how the CAF DTD is used, Figure 4 shows a CAF XML file with one domain map only, called ‘Questions’, with sub-concepts titled ‘Question1’ and so forth. All concepts have a title attribute and the questions have each a question and an answer attribute. The goal model part always has one lesson (since CAF only describes a lesson and the Domain Maps on which it is based). In this example, the ‘lesson’ (goal map) is comprised of the above mentioned questions, in the exact same order and structure as in the original domain map. Note that in the goal map, labels (‘question’ and ‘answer’, respectively) have been added for the questions and answers.
42 <CAF> <domainmodel> <! ...> <concept> <name>Questions</name> <! ...> <concept> <name>Question1</name> <attribute> <name>question</name>
<contents>Where is Hypertext 2008? </contents> </attribute> <attribute> <name>answer</name> <contents>Pittsburgh</contents> </attribute> </concept>
< ! ... following concepts omitted > </concept>
</domainmodel>
<goalmodel>
<! ...>
<lesson weight="0" label="question">
<link weight="0" label="question"> Questions\Question1\question </link> <link weight="0" label="answer"> Questions\Question1\answer </link> </lesson>
<! ... following lesson parts omitted> </goalmodel>
</CAF>
43 2.6.2 Ontology Based Content Description
Authoring content for AH, free from adaptation, is closely related to modelling the specific subject domain conceptually. In this section we introduce ontology based methods of describing the domain content.
The Adaptive Course Construction Toolkit (ACCT, see also section 2.7.4) [50], [51] is an authoring system for AH which is based around services and activities for learning. ACCT is an attempt to combine the best of the activities and learning-design driven authoring, such as in the Learning Activity Management System (LAMS) [107] system, and content driven AH authoring methods, such as represented by the AHA! Graph author. The system describes the domain it is teaching conceptually in a so called Subject Matter Concept Space (SMCS). This is a light-weight ontology describing the relationships within a subject matter domain. The ACCT supports the authoring of this Concept Space. It allows the definition of relationships between the concepts. The relationships are provided as patterns that can be customised.
As we will see in chapter 6, this is highly related to the description of the Domain Model in our novel graphical authoring model, the CAM.