Cik: The k-th associated concept of the i-th NEA node Factnow: The name of current node being studied.
Eij: The edge from Ni to Nj
FactNj: The next node to be studied, if Nj is satisfied.
FactCik: The score of Cik
<OP>: The relational operator, i.e., =, >, <, ≦, ≧ and ≠ Input: The XML file of OOLA model
Output: The XML file with a Rule Class of DRAMA Inference Engine Step 1: Create a Factnow
Step 2: For each Eij∈OOLA model
Step 3: Output the XML file of DRAMA
Chapter 6 Knowledge Manager (KM)
In e-learning system, teaching materials are usually stored in database, called Learning Object Repository (LOR). Because the SCORM standard has been accepted and applied popularly, its compliant teaching materials are also created and developed.
Therefore, in LOR, huge amount of SCORM teaching materials including associated learning objects (LO) will result in the issues of management. Recently, SCORM international organization has focused on how to efficiently maintain, search, and retrieve desired learning objects in LOR for users. Therefore, in this dissertation, , we propose a new approach, called Level-wise Content Management Scheme (LCMS) [116], to efficiently maintain, search, and retrieve the learning contents in SCORM compliant LOR and it is implemented within a Learning Object Repository (LOR) Manager in Knowledge Manager (KM) of ILCMS.
6.1 Level-wise Content Management Scheme (LCMS)
6.1.1 The Processes of LCMS
As shown in Figure 6.1, the scheme of LCMS is divided into Constructing Phase and Searching Phase. The former creates the content tree form SCORM content package by CP2CT process, and then creates and maintains a multistage graph as Directed Acyclic Graph (DAG) with relationships among LOs, called Level-wise Content Clustering Graph (LCCG), by applying clustering techniques. The latter traverses the LCCG by LCCG Content Searching Algorithm (LCCG-CSAlg) to retrieve desired learning content with general and specific LOs according to the query of users over wire/wireless environment.
Constructing Phase includes the following three processes:
z Content Package to Content Tree (CP2CT) Process: it transforms the content structure of SCORM teaching materials (Content Package) into tree-like structure with the representative feature vector and the same depth, called Content Tree (CT), for representing each teaching materials.
z Level-wise Content Clustering Process: it clusters LOs according to content trees (CTs) to establish the level-wise content clustering graph (LCCG) for creating the relationships among LOs .
z LCCG Maintaining Process: it monitors the condition of each node within LCCG and to rebuild the LCCG if necessary.
Searching Phase includes the following two processes:
z SCORM Metadata Searching: it first searches the desired whole teaching materials by the associated SCORM metadata for addressing the related nodes as entries of LCCG.
z Level-wise Content Searching: it then traverses the LCCG from these entry nodes to retrieve the more precise learning objects in LOR and to deliver these for learners.
Figure 6.1: The Flowchart of Level-wise Content Management Scheme (LCMS)
6.1.2 Content Package to Content Tree (CP2CT) Process
Because we want to create the relationships among LOs according to the content structure of teaching materials, the organization information in SCORM content package will be transformed into a tree-like representation with representative feature vector, called Content Tree (CT). For clustering process conveniently, the depth of every CT is the same. Its definition is described as follows.
Definition 6.1: Content Tree (CT) = (N, E), where z N = { n0, n1,…, nm }.
z E = { nini+1 | 0≦i< the depth of CT }.
In CT, each node is called “Content Node (CN)” containing a feature vector V which denotes the representative feature of learning contents within this node. E denotes the link edges from node ni in upper level to ni+1 in next lower level.
In this dissertation, we apply the Vector Space Model (VSM) approach [24] [96] to represent the learning contents in CN. Thus, based upon the Term Frequency - Inverse Document Frequency (TF-IDF) weighting scheme [2] [30] [68] [106] [136], each CN can be represented by an N dimensions vector as <tf1×idf1, tf2×idf2,…, tfn×idfn>, where tfi is the frequency of the i-th term (keyword) and idfi=log(m/df(t)) is the Inverse Document Frequency (IDF) of the i-th term in the document (where m is total number of documents and df(t) is the number of documents that contains the term).
For conveniently creating the relationships among learning objects according to the content structure, we assume that every content tree (CT) transformed from content package will have the same depth of tree. However, in many teaching materials, the depths of content structures are different. Therefore, in CT, if the depth of a leaf CN is too short, the Virtual Node (VN) will be repeatedly inserted as its child node until the difference of the desired depth has been filled. The feature vector of every VN is the same as its parent CN or VN. Besides, if the depth of a leaf CN is too long, its parent CN in the desired depth will merge the information of all included child nodes into one new CN whose feature vector is generated by averaging these included child nodes.
The Example 6.1 shows the process of transforming the organization information of SCORM content package into Content Tree (CT) with the feature vector V and the same depth.
Example 6.1:
Given a SCORM content package shown in the left side of Figure 6.2, we take TF-IDF as weighting scheme to create the feature vector V in each CN node. Because the depth of CN, “Chapter 1”, is too short, the VN named “1.1” is inserted and its feature vector V =<3, 2, 2> is the same as 21 V . Moreover, the CN, “3.1”, is too long, 11
so that its included child nodes, i.e., “3.1.1” and “3.1.2”, are merged into one CN, “3.1”, and their feature vector V is the average of <1,0,1> and (<2,1,0>+<0,1,2>)/2 after 24
the rolling up process. Then, The CT after CP2CT Process is shown in the right part of Figure 6.2.
Figure 6.2: The Corresponding Content Tree (CT) of the Content Package (CP) by CP2CT process