符合教學需求之適性化學習環境
10
0
0
全文
(2) At present, SCORM 1.3 adopts the Simple Sequence Specification (SSS) of IMS [3] to define the course sequencing behavior. The SSS relies on the concept of learning activities. A learning activity may be loosely described as an instructional event or events embedded in a content resource, or as an aggregation of activities to describe content resources with their contained instructional events. Content in simple sequencing is organized into a hierarchical structure, namely Activity Tree (AT). The Simple Sequencing process uses information about the desired sequencing behavior to control the sequencing, selecting and delivering of activities to the learner. Therefore, by this standard, we can develop an intelligent approach to (semi-)automatic course or exercise sequencing. Simple Sequencing makes no requirements on the structure, organization or instruction of the Activity Tree. The tree and the associated sequencing definitions may be static or dynamically created. However, how to create, represent and maintain the Activity Tree and associated sequencing are our concerns. For a large learning activity, the Activity Tree will become too complex to manage and reuse, much less its sequencing behavior rules. Moreover, the lack of inter-relations among Activity Trees also makes the reusing and the reintegrating hard. These defects will limit the scalability and flexibility of adaptive learning system. Existing intelligent tutoring systems often use the grade and learning duration of learner to evaluate the learning result and to decide the delivering sequence of learning content. However, to evaluate the personal learning behavior, the grade and learning duration may be not sufficient. Therefore, many researches used Pedagogical Theory [14][17][20][22]to enhance the evaluation of the personal learning characteristic and to generate more personalized learning guidance. Hence, in this paper, the learning characteristic of learner has been taken into consideration. Our approach is to extend and modualize the structure of AT by applying Pedagogical Theory and concept of Object Oriented Methodology respectively to construct a personalized learning environment. One large AT is modualized into several suitable AT nodes which possess several specific attributes and associated inter-relations. By these properties, each AT node can be reused and reintegrated to generate new organization of course for decreasing the cost of designing learning activity, and increasing the scalability and flexibility. For the needs of specific purpose, e.g. pedagogical needs, the attribute and interrelation of AT can be extended, too. Therefore, we propose a novel model, Instructional Activity Model (IAM), which is composed of related Activity Tree nodes. Based upon Pedagogical Theory, each. AT node in IAM is defined as a learning unit with inter-relations among AT nodes and specific attributes which the IAM can be easily managed, reused, and integrated by. In addition, we also propose two heuristic algorithms with general and specific domain heuristic to traverse IAM to generate the dynamic learning content to the learner. IAM with scalability and flexibility can apply different pedagogical theories for specific need by extension mechanisms. In addition, based on SCORM 1.3 RTE (Run Time Environ ment) [2], we develop an IAM system to manage and dynamically generate personalized SOCRM compliant course. Based upon the results of experiment, we may conclude that the IAM system is workable and compatible with SCORM standard. The main contributions of this paper are: 1. Propose a general purposed model, called Instructional Activity Model (IAM), to generate adaptive learning course which is compatible with SCORM standard. 2. Modualize a large Activity Tree into several suitable AT nodes with specific attributes and inter-relations which can be easily managed, reused, and reintegrated, based on OO methodologies. 3. Apply Pedagogical Theory in IAM to define the personal learning characteristic and evaluate learning result of learner for generating more adaptive learning. 4. Develop an IAM system, which integrates Simple Sequencing Specification, inference engine, DRAMA, and SCORM RTE 1.3, to deliver individual learning course.. 2. Related Work In this section, we review SCORM standard and some related works as follows. 2.1 SCORM (Sharable Content Object Reference Model) [2] Among those existing standards for learning contents, SCORM is currently the most popular one. It is a product of the U.S. Government's initiative in Advanced Distributed Learn ing (ADL). In November of 1997, the Depart ment of Defense and the White House Office of Science and Technology Policy launched the ADL initiative with the goal of providing access to high-quality education and training materials that are easily tailored to individual learner needs and available whenever and wherever they are needed. The SCORM specifications are a composite of several specifications developed by international standards organizations, including the IEEE, IMS, AICC and ARIADNE..
(3) In a nutshell, SCORM is a set of specifications for developing, packaging and delivering high-quality education and training materials whenever and wherever they are needed. SCORM-compliant courses leverage course development investments by ensuring that compliant courses are "RAID:" Reusable: easily modified and used by different development tools, Accessible: can be searched and made available as needed by both learners and content developers, Interoperable: operates across a wide variety of hardware, operating systems and web browsers, and Durable: does not require significant modifications with new versions of system software [8]. 2.2 Simple Sequencing Specification [2][3] The Simple Sequencing specification of IMS [3], which was adopted by SCROM 1.3, relies on the concept of learning activities. A learning activity may be loosely described as an instructional event or events embedded in a content resource, or as an aggregation of activities to describe content resources with their contained instructional events. Content in simple sequencing is organized into a hierarchical structure, namely activity tree (AT) as a learning map. The exa mples of AT are shown in Figure 1. Each activity including one or more child activities has an associated set of sequencing behaviors, defined by the Sequencing Definition Model which is a set of attributes used by Simple Sequencing. The Simple Sequencing process uses information about the desired sequencing behavior to control the sequencing, selecting and delivering of activities to the learner. The intended sequence is described by a specific set of data attributes, which are associated with learning activities in the activity tree to describe the sequencing behavior. Moreover, the activity tree can be considered as the learning map. The sequencing behaviors describe how the activity or how the children of the activity are used to create the desired learning experience. Simple Sequencing makes no requirements on the structure, organization or instruction of the activity tree. The tree and the associated sequencing definitions may be static or dynamically created. Therefore, how to create, represent and maintain the activity tree and associated sequencing definition, which is not specified, is an important issue. Simple Sequencing enables us to share not only learning contents, but also intended learning experiences. It provides a set of widely used sequencing method so that the teacher could do the sequencing efficiently. However, the definition of sequencing behavior rules is obviously too simple to satisfy pedagogical needs. 2.3 Other Related Research Carchiolo [12] has proposed adaptive formative paths for e-learning environment. They con-. structed domain database and students’ profiles to obtain personalized learning paths. During the learning process, the learning paths can be dynamically modified according to student needs and capabilities. Although this system has some advantages including consideration of student’s priori knowledge and generation of adaptive learning path, it didn’t take pedagogical theory into account and it could not be compatible with SCORM standard yet. Sheremetov [19] also proposed a system, called EVA, for developing a virtual learning space in the National Technical Institute in Mexico. EVA consists of five virtual learning spaces: 1. Knowledge Space: all necessary in formation to learn, 2. Collaborative Space: real or virtual companions that get together to learn, 3. Collaborative Space: the teachers or tutors (also real and virtual) who give the right direction for learning and consult doubts, 4. Experimentation Space: the practical work of the students in virtual environment to obtain, and 5. Personal Space: records about users’ information. The model of knowledge is represented in the form of graph, and each node is a unit of learning material (ULM) which is the basic element of knowledge structure. ULMs with a related knowledge concept can be grouped into a POLIlibro (or Multi-Book) along the learning trajectory (path) depending on students. However, the relations between ULMs are not sufficient for expressing the structure of the knowledge model and the attributes of an ULM is not enough for mining the behaviors of students. They also proposed some methodologies for planning of trajectories and scheduling of learning activities based on the agent technology. However, they didn’t discuss how to generate a learning path.. Figure 1: Examples of Activity Tree. 3. Instructional Activity Model (IAM) As mentioned above, SCORM standard defines a hierarchical structure, namely Activity Tree (AT), to sequence the learning content to the learner. By defining the sequencing behavior rules within AT, we can develop an intelligent approach to (semi-)automatic course and exercise sequencing. However, how to create, represent and maintain the Activity Tree and associated sequencing definition are our concerns. For a large learning activity, the Activity Tree will become too complex to manage and reuse. It is hard to reuse and.
(4) reintegrate ATs without the inter-relations. These defects will limit the scalability and flexibility of adaptive learning system. Moreover, for more personalized learning, many researches have used the Pedagogical Theory [14][17][20][22] to enhance the evaluation of the personal learning characteristic. Hence, in this paper, we extend and modualize the structure of AT by applying Pedagogical Theory and concept of Object Oriented Methodology, respectively. Thus, we propose a novel model, Instructional Activity Model (IAM), which is composed of related Activity Tree nodes. Based upon Pedagogical Theory, each AT node in IAM is modualized as a learning unit with in ter-relations and specific attributes, which can be easily managed, reused, and integrated. We also propose two heuristic algorithms with general and specific domain heuristic, to traverse IAM to generate the dynamic learning content to the learner. In this section, we will describe the concept of Instructional Activity Model including its properties and the learning process algorithms. 3.1 Concept of Instructional Activity Model In Simple Sequencing Specification (SSS), we can create an AT on the fly. As mentioned above, in a large AT, its organization and sequencing rules definition are hard to manage and reuse. However, too many small ATs also result in that the management of AT nodes and rules definition is complicated. Therefore, for the scalability and flexibility of AT, how to define a suitable unit of AT becomes an important issue. According to Bloom’s Mastery Theory [10], a suitable unit of learning content is a chapter or a section for learning. Thus, in IAM we define the unit of an AT as a chapter or a section. Assume there are n ATs. We define an AT set as ATset={AT1 , AT2 ,… ,ATn }. According to the formulation of Gagne [15]: “A capability is a knowledge unit stored in a person’s long term memory that allows him/her to succeed in the realization of physical, intellectual or professional activity.” Therefore, to suppose there are m capabilities, we can also obtain Cset={c1 , c2 ,… ,c m}. Before learning an AT, students are supposed to possess some capabilities, called Prerequisite. Similarly, after learning an activity tree, students can acquire further capabilities, called Contributions. Every prerequisite and contribution has its own weight representing significance before and after learning. Therefore, in IAM the Cset can be regarded as the union of all prerequisites and contributions and an AT has several capabilities. A learning activity or a course is composed of several ATs with input/output capabilities. The student will learn easy AT he can qualify and gain further capabilities which enable student to learn another advanced AT. This learning process will be repeated until the student has finished all learning objectives. Then, every student will have. an individual value of Cset. The Figure 2 shows the diagram of IAM.. Figure 2: The Diagram of IAM In Figure 2, we can define these terms as follows: l eij : is an edge from c i to ATj , called “prerequisite edge”. It means that before learning ATj , the student is supposed to possess this ability ci . l e'ij : is an edge from ATi to cj , called “contribution edge”. It means that after learning ATi , the student will gain the ability cj . l w(e ij) : is the weight of eij . It represents the significance of ci before learning ATj . Note that the sum of w(e ij) of an AT is 1, that is, ∑ w(eij ) = 1, ∀j . i. l w(e'ij) : is the weight of e’ij . It represents the significance of cj after learning ATi . Note that the sum of contribution weights of an AT is 1; that is, ∑ w(e'ij ) = 1, ∀i . j. l mReqij : is the minimum requirement of ci for learning ATj , which is denoted as mReqij . We use it to determine whether the student is qualified to learn ATj or not. l mFuni (AT) : is the measure function of ATi . It is used to evaluate whether the student is qualified to learn ATi or not. We can define Measure Function (mFun) for each AT; e.g., the mFun of AT2 can be defined as mFun2 = val(c2 )×w(e 22 )+ val(c3 )×w(e 32 ). l grade(e'ij) : represents the learning grade after learning ATi . l val(c m ) : the value of capability m (cm ). ∑ w(e'j m ) × grade(e'j m ) val(c m) = j , w(e' ) ∑ jm j. l Student’s Acquired Capability (AC): records his/her learning results in the form of AC = U (ci , val(c i )) , l Student’s Course Objectives (CO): records his/her learning objectives in the form of CO = U ci . l Potential Capability List (PCL): Each AT has a potential capability list recording all the contribution capabilities which can be reached from this AT via edges in IAM. It can be formulated as PCLATj = U ci , //Ci can be reach.
(5) form ATj by connected edges//. For example, in Figure 2 the PCLAT1 equals {C2 , C3 , C4 , C5 , C6 }. l Student’s Grade Prediction (SGP): represents the performance prediction of the specific student related to the activity tree. We use measure function (mFun) and acquired capabilities (AC) to compute SGP value of each AT, For example, SGP 2 =mFun2 =val (c2 )×w(e22)+ val(c3)×w(e32) l Normalized Objective Weight (NOW): represents the relativity between an activity tree and the student’s course objectives. Higher objective weight implies better learning efficiency. Empirically, selecting function tends to select the activity tree with higher SGP and NOW for students. We propose two following approaches to compute the value of NOW. 1. Normalized Objective Weight Computing by Heuristic1 (NOWCh1 ): NOW. h1 the number of c i (c i ∈ PCL & c i ∈ CO) = j the number of c i (c i ∈ PCL). 2. Normalized Objective Weight Computing by Heuristic1 (NOWCh2 ): Step1: ∀ci in PCLj , use DFS to find the path between ATj and ci , pathi is of the form e’j1 , e12 ,… , e’ni , where each set is a directed link in IAM Step2: ∀ci in PCLj , compute its path donation (PDi ) where PDi = product of every link donation (LD) = LD(e’j1 ) × LD(e 12 ) ×… × LD(e’ni ) where w ( e' xy ) if link is of the form e' xy , LD(e' xy ) = w (e' iy ) ∑ i if link is of the form e xy , LD(e xy ) = w(e xy ). Step3:. ∑ PD. i. NOW. h2 j. =. i, c i ∈ CO. ∑ PD. i. i. l Chosen Factor (CF): is a linear combination of selecting criteria, NOW and SGP. CFi = aNOW i + ßSGPi , where a+ß=1, 0≤ a, ß ≤1. In brief, Instructional activity model (IAM), a graphical based representation of a learning activity or course, contains several ATs, Capabilities including prerequisite and contribution, Relations Edges including eij with mReqij and e’ij , and several Measure Functions. It can be formulated as IAM=(ATset, Cset, Eset , E'set), where l ATset = {AT1 , AT2 , … ,ATn }. l Cset = {c1 , c2 ,… ,cn }. l Eset is the set of all prerequisite edges with minimum requirement value in an IAM. It is formulated as Eset = U Ej , Ej = (eij, mReqij ) ,. U i. eij ∈ATj .. l E'set is the set of all contribution edges in an IAM. It is formulated as E'set = U E' j , E'j =. U e'. jk. , e'jk ∈ATj .. j. Figure 3: The Flowchart of Learning Process 3.2 Basic Functionalities Based upon the structure of IAM described above, we can develop several approaches to provide students a learning environment with a dynamic and adaptive course. The learning process can be simply considered as sequencing activity trees in IAM in order to let students satisfy learning objectives. The flowchart and algorithm of learning process is shown in Figure 3 and Figure 4, respectively. Here, we explain the learning process algorithm of IAM. First, we initialize the learning state by loading AC and CO, and evaluate the PCLAT of every AT (Setp1-Step2), and then enter the loop of learning activity (Step3). During learning process, we mark each AT with Candidate tag or Blocking tag by comparing the mReq(e ij ) with val(c i ) (Step3.1). The Candidate type represents that this AT will be selected and the Blocking type is opposite just. Then, we select a suitable AT with candidate type first and deliver it to the learner (Step3.2). If there exists no AT with candidate, the learning process goes into the Re medy Course Process (Step3.2.1-Step 3.2.5). At the Remedy Course Process, we will select an AT with the largest amount of Cm∈CO (Step3.2.1) and then find a ci with the smallest or largest value of (mReq(e ij )－val(c i ) according to SelectingPolicy (Step3.2.2-Step3.2.3). In this algorithm, we provide three policies to select different capability for adaptive learning. The policy of “Easiest First” trends to select a ci which the learner has gained more high grade in but policy of “Hardest First” is opposite. After selecting a ci , we can decide which AT connected with ci to deliver to learner by computing MAX((mReq(e ij ) – grade(e'jk )) * w(e'jk )) which implies that the progress of learner is the largest (Step3.2.4). When learner has finished and satisfied all course objectives, the learning process will be stopped..
(6) Table 1: The Related Value of AT1 and AT2 SGP NOWh1 AT1 val(c 1 ) the number of { c 4 , c 7 , c 8 } 3 × = = 0.6 the number of { c 4 , c 5 , c 7 , c 8 , c 9 } 5 w(e 11 ) =0.82×1 NOWh2 PD + PD + PD =0.82. AT2. SGP2 =0.45. 4. + PD. 4. 7. 8. + PD 7 + PD 8 + PD 9 1 + (1 × 1 × 0 . 67 ) + (1 × 1 × 0 . 25 ) = 1 + 1 + (1 × 1 × 0 . 67 ) + ( 1 × 1 × 0 . 25 ) + ( 1 × 0 . 7 × 0 . 56 ) = 0 . 46 NOW h1 2 =0.33 NOW h2 2 =0.152 PD. 5. Algorithm 1: Learning Process Algorithm Input: IAM: instructional activity model, AC: acquired capabilities with normalized grade of student CO: course objectives of the student. SelectingPolicy:{Easiest First, Medium First, Hardest First} Output: AC: new AC after learner has finished learning activity. Algorithm : Step1: Load the learner’s Acquired Capabilities (AC) and Course O bjectives (CO). Step2: Evaluate all AT ∈IAM with PCL AT Step3: while(CO⊄AC) //start the learning activity Begin // decide type of AT is Candidate or Blocking state 3.1: for all Ci with eij in AC { if (mReq(e ij)>val(ci)) then mark the AT j with Blocking else if (AT j has not been learned yet) then ( compute CFj) and (mark the AT j with Candidate) } //select a suitable AT to be learned 3.2: if (∃ any AT with Candidate mark) // select the AT with Candidate mark then (Select an ATj with the highest CFj and delivery it to the learner.) else if (∃ any AT with Blocking mark) then//go to Remedy Course Process & select a suitable AT 3.2.1: for all AT j {Count the amount of Cm∈CO which is connected by e'jm.} 3.2.2: Select the AT j with the largest amount of Cm∈CO. 3.2.3: for all ci with eij { if SelectingPolicy=”Easiest First”, then Find ci with the smallest (mReq(eij)－val(ci)). else i f SelectingPolicy =”Hard est First”, then Find ci with the largest (mReq(eij)－val(ci)). else i f SelectingPolicy =”Medium First”, then Find ci with the medium (mReq(eij)－val(ci)).} 3.2.4: Select AT k with ∀i, MAX( (mReq(eij)–grade(e'jk ))*w(e'jk)), e'jk ∈Ej. 3.2.5: Clear mark of AT j and delivery AT k to learner. 3.3: if learner pass the selected AT then mark this AT with Learned. 3.4: update AC after the learner learns selected AT. End Step4: return AC.. Figure 4: Learning Process Algorithm SelectingPolicy (Step3.2.2-Step3.2.3). In this algorithm, we provide three policies to select different capability for adaptive learning. The policy of “Easiest First” trends to select a ci which the learner has gained more high grade in but policy of “Hardest First” is opposite. After selecting a ci , we can decide which AT connected with ci to de-. CFh1 α × SGP. 1. + β × NOW. h1. = 0 . 5 × 0 . 82 + 0 . 5 × 0 .6 = 0.7. CFh2. 0 .5 × 0 .82 + 0.5 × 0 .82 = 0 .64. CF 2h1 = 0 . 39 CF 2h 2 = 0 . 301. liver to learner by computing MAX((mReq(e ij )–grade(e'jk ))*w(e'jk )) which implies that the progress of learner is the largest (Step3.2.4). When learner has finished and satisfied all course objectives, the learning process will be stopped. Example 1: This IAM in Figure 5 can be represented as follows: IAM = ({AT1 , AT2 AT3 , AT4 , AT5 ,}, {c1 , c2 , c 3 , c4 , c5 , c6 , c7 , c8 , c9 }, {(e 11 ,0.8), (e 22 ,0.7) , (e 23 ,0.8), (e 33 ,0.8) , (e 44 ,0.8) , (e 55 ,0.8) , (e 65 ,0.6)}, {e'14 , e'15 , e'25 , e'26 , e'47 , e'48 , e'58 , e'59 }). Case1: We assume that AC={c1 (0.82), c2 (0.75) } and CO={ c4 , c7 , c8 }. Note that the value in parenthesis is the val(c i ). The PLCAT has been evaluated as shown in Figure 4. After first iteration of the While loop of algorithm, we can get results as shown in Table 1. Thus, AT1 will be delivered to learner because it has highest CF value.. Figure 5: The Example of IAM Case2: we assume that AC={ c1 (0.82), c2 (0.75) , c4 (0.75) , c5 (0.6) , c6 (unknow) }, CO={ c4 , c7 , c8 }, and Blocking AT = {AT3 , AT5 }. The learning process has gone into Remedy Course Process. Before the Step 3.2.3, because AT5 has one cm ∈ CO, AT5 is selected. If the SelectingPolicy is “Easiest First”, the C5 with the smallest value 0.2 of (mReq(e 55 ) － val(c 5 )) is selected. Then, by computing the (mReq(e 55 )–grade(e'15 )) × w(e '15 )).
(7) and (mReq(e 25 )–grade(e'25 )) × w(e '25 )), we can decide to deliver the AT2 with value 0.26 to learner.. 4. Applying Pedagogical Theories of IAM We have proposed a n Instructional Activity Model with several algorithms to provide scalability and flexibility of course sequencing in a dynamic way. This general model can be specialized by utilizing pedagogical theories. Now we are going to map several well-known pedagogical theories to IAM using extension mechanisms. 4.1 Pedagogical Theories We consider pedagogical theories in three parts. In each part, we apply a well-known pedagogical theory to IAM in an attempt to show its applicability in practice. The three parts are 1. Organization of IAM, 2. Learning style of students, and 3. Capability taxonomy, which are described as follows. 1. Organization of IAM It is essential to organize suitable teaching materials for students. According to the proposal of Nelson L. Bassing [9], we can categorize organizations of teaching materials into three types: 1).logical organization: e.g., teaching from basic idea to advanced idea in mathematics, or teaching from ancient times to modern times in history, sequences teaching materials in a systematical fashion, 2).psychological organization: places emphasis on the student’s own in terest, ability, and needs, and 3).eclectic organization: does not take side with either logical organization or psychological organization but takes both of them into consideration. 2. Learning style of learner There are lots of classification taxonomies collected in [21] based on different criteria. However, most of them are not designed for e-learning environ ment. We will simply utilize VAK learning style in IAM. V implies visual, A implies auditory, and K implies kinesthetic. Existing questionnaire [21] is used to extract learners’ learning style. After each learner finishes the questionnaire, we can obtain his sensibility in visual level, auditory level, and kinesthetic level. We utilize this information in activity tree selection. 3. Capability taxonomy Capabilities in IAM can be classified into several types based on different needs. We use what proposed by Gagne [14] to do the classification. Gagne considers that learning outcomes can be classified into five types: verbal information, intellectual skills, cognitive strategies, motor skills , and attitude. We will categorize capability into five types. based upon Gagne’s theory. Related defin itions and modifications will be discussed in this section. 4.2 Definitions and Properties We are going to modify definitions corresponding to the model we proposed according to those pedagogical theories. First we categorize capability into five types: verbal information, intellectual skills, cognitive strategies, motor skills , and attitude. Suppose there are n capabilities in a specific domain, each capability is one of the five types. We obtain Cset by grouping these capabilities together. Cset = {c1 , c2 ,… cn }, where each ci has five dimensions: <vci , ic i , cci , mci , aci >, and vci denotes verbal capability, ic i denotes intellectual capability, cci denotes cognitive capability, mc i denotes motor capability, and aci denotes attitude capability. With this new definition of Cset, our proposed algorithms are capable of searching five types of capabilities without further modifications. We give a few definitions for pedagogical need as follows. Definition 1: LnStyi of ATi Each activity tree is accompanied with LnSty as a vector to represent its learning style in visual degree, auditory degree, and kinesthetic degree. Each value in this vector is between 0 and 1. Definition 2: LgOrg i of ATi LgOrg, e.g. the default teaching order of the activity tree, indicates the logical organization of an activity tree. Teaching order is labeled in consecutive number in traditional course. We discretize the teaching order and map it to interval [0,1]. For example, LgOrg of activity tree is either one of {0.3, 0.6, 0.9}. Definition 3: Student’s learning style (SLS) We create a vector, called student’s learning style (SLS), for each student to characterize his/her learning style. Step 3.1: … else if (AT j has not been learned yet) then (compute CFj(1) and CFj(2)) and (mark the AT j with Candidate) … where: CFi(1) = α NOW i + ß SGPi + γ LgOrg i , α , β , γ are constants. (2) CFi = SLS • LnSty i. Step 3.2: i f (∃ any AT with Candidate mark) // select the AT with Candidate mark then 1: sort CFj(2) in descending order, delete AT whose CFj(2) is in the second half. 2: Select an ATj with the highest CFj and deliver it to the learner. … .. Figure 6: The Pedagogical Learning Process Algorithm The first element represents level of visual characteristics, the second one represents level of audio characteristics, and the third one represents.
(8) level of kinesthetic characteristics. We utilize this vector in activity tree selecting function to determine if the activity tree suits the student’s learning style.. Figure 7: An Example of IAM with Pedagogical Theories. 4.3 Learning Process with Pedagogical Theories We are going to make a few modifications to accommodate the algorithms to pedagogical theories. The modified learning process algorithm has only slight difference than the previous one. We use learner’s learning style as a criterion for selecting suitable activity tree for student’s learning. 4.3.1 Learning Process Algorithm with Pedagogical Theories In the section 3.3, Learning Process Algorithm selects activity tree based on SGP and NOW, now we propose a Pedagogical Learning Process Algorithm, which is extended form Learning Process Algorithm, to select AT by the criteria of pedagogical theories, i.e. logical organization and psychological organization. From our vie wpoint, logical organization is characterized by the order of activity trees and psychological organization is characterized by student’s learning style, grade prediction and course objectives. The four characteristics are used as criteria for selecting AT. Pedagogical Learning Process Algorithm is to extend the Step3.1 and Step3.2, which are shown in Figure 6, of Learning Process A lgorithm. In Step 3.1, we compute CFj (1) and CFj (2) to represent characteristics of psychological organization and logical organization. We regard these two as partially, linearly dependent. Normalized Objectives Weight (NOW), Student’s Grade Prediction (SGP), and logical organization are mutually linearly dependent and can be computed by a simple linear function. Learning style, non-linearly dependent with the three, is computed by product of student’s learning style vector (SLS) and activity tree’s learning style vector (LnSty). The more similar these two vectors are, the higher the product is. Step 3.2 represents a selecting strategy, first we sort CFj (2) in descending order and prune the latter half. Then we select the activity tree whose CFj (1) is the highest. This strategy implies we consider psychological organization first and prune those invalid activity trees, and then we consider the physical organization to select the most suitable activity tree.. Table 2: Learning style and logical organization of each AT. AT1. AT2. AT3. LnSty LgOrg. <0.8 , 0.1 , 0.1 >. <0.1, 0.8 , 0.1 >. <0.6 , 0.1 , 0.3 >. 0.3 AT4. 0.3 AT5. 0.5. LnSty LgOrg. <0.2 , 0.1 , 0.7 >. <0.1 , 0.2 , 0.7 >. 0.3. 0.7. Since pedagogical objectives differ from learners, courses, teachers , etc., we can modify Step 3.1 to Step 3.2, if needed, to fulfill pedagogical needs. Different selecting strategies generate different sequencing results. Example 2 : Learning in IAM with pedagogical theories We are going to show a simple example of learning in IAM with pedagogical theories. First we define IAM and related attributes of each activity tree; thereafter we’re going to demo nstrate the process of Learning Process Algorithm for a specific student. The exa mple of IAM is shown in Figure 7. IAM in Figure 7 is represented as follows : IAM=({AT1 , AT2 , AT3 , AT4 , AT5 }, {vc1 , cc2 , mc3 , vc4 , ic 5 , vc6 , mc7 , ic 8 , cc9 , ic 10 }, {(e 11 ,0.3), (e 12 ,0.6), (e 22 ,0.5), (e 33 ,0.4), (e 44 ,0.5), (e 55 ,0.6), (e 65 ,0.5)}, {e’14 , e’15 , e’25 , e’26 , e’36 , e’37 , e’48 , e’49 , e’5,10}) The learning style and logical organization, being used in Pedagogical Learning Process Algorithm, are shown in Table 2. Note that the logical organization (LgOrg) of each activity tree is represented in numerical format. Numbers 0.3, 0.5 and 0.7 can be regarded as basic, medium, and advanced teaching resources, respectively. Suppose there’s a learner learning in this IAM, his/her personal information is AC={(vc1,0.5),(cc1,0.8),(mc1,0.1),(vc2,0.6),(ic2,0.6)}, SLS =<0.1, 0.2, 0.7>, CO={(vc3,0.3), (ic2,0.4), (ic3,0.6), (ic4,0.5), (cc2,0.6), (mc2,0.5)}.. Since he/she had learned AT1 , Learning Process Algorithm will choose the next activity tree for his/her learning. CFi (1) and CFi (2) in Pedagogical Learning Process Algorithm are defined as CFi (1)=0.2×NOW i +0.2×SGPi +0.4× LgOrg i , CFi (2)= SLS • LnStyi , where the symb ol “ • ” represents the dot product. To execute Learning Process Algorithm, we can compute NOW, SGP, CFi (1), and CFi (2) and other information as shown in Table 3. Table 3: Selecting Criteria for Each Activity Tree. AT2 AT3 AT4 {ic2,vc3,ic4} {mc2,vc3,ic4} {ic3, cc2} PCL 1 1 1 NOW h1 0.5×0.4+ 0.8×0.6=0.68 0.3 LgOrg 0.24 SLS • LnStyi 0.456 CFi (1) 0.24 CFi (2). SGP. 0.1×1=0.1. 0.6×1=0.6. 0.5 0.29 0.42 0.29. 0.3 0.53 0.44 0.53.
(9) Then, we use the following strategy for selection: for smart students, select the AT with the highest CFi (1) value; for remaining students, select the AT with the highest CFi (2) value. With this strategy, we select AT2 for smart students, AT4 for remaining students. In addition, we can revise CFi (1) and CFi (2) for specific purposes. For exa mple, some teachers believe that learning style have relations with student’s grade, they can modify CFi (1) and CFi (2) as CFi (1) = 0.5 × NOW i + 0.5 × LgOrg i , CFi (2) = 0.5 × SGPi + 0.5 × (SLS • LnStyi ) . If the selecting strategy remains the same, we will provide AT3 for smart students, and AT4 for remaining students. 4.4 Estimation of the Power of IAM We have shown that it is possible to apply pedagogical theories to our IAM for specific need. However, how many pedagogical theories can be applied to IAM? In this section, we estimate that how many different structures IAM can support for pedagogical needs. Educational researchers proposed various types of course structure to facilitate student’s learning. Posner [18] proposed three types of structures including discrete structure, linear structure, and hierarchical structure. Bruner [11] proposed concept of spiral curriculum. And Efland [13] proposed the lattice curriculum. Each structure satisfies certain kinds of pedagogical need. IAM we proposed can be applied to those course structures, as can be seen in Figure 8 and Figure 9.. Figure 8: IAM Mapping to Discrete Structure, Linear Structure, and Hierarchical structure. AT Tree Inference Engine Module includ ing AT Parser and DRAMA Inference Engine. In IAM Module, the IAM Controller mainly manages the IAM Model and communicates with Controller and AT Tree IE Module. While learner logins system, IAM system will assign a Controller, which will manage all events during learn ing, to learner. In AT Tree IE Module, the Inference Engine, called DRAMA, is the Object-Oriented Rule Base Management System (OORBMS) which was proposed based on Drama Model (DM) for constructing a reusable, sharable, and modifiable knowledge base [1][23][24][25]. The model manages rules under object-oriented paradigm. Moreover, the rule format of Simple Sequencing Specification possesses the general rule format as: if condition set Then action. This format can make us to easy adopt Drama Model to generate the sequencing rule set. During learning, the learner requests a course to learn and then the SCORM RTE 1.3 will send this request to the IAM Controller via Controller. The IAM Controller acquires learner’s information and personal IAM Model and then selects an AT to inform AT Parser to extract and deliver rules to DRAMA Inference Engine. Finally, SCORM RTE receives the inference result of DRAMA and then deliver suitable course to learner.. Figure 10: The Architecture of IAM System based on SCORM RTE 1.3. 6. Conclusion. Figure 9: IAM Mapping to Spiral Curriculum and Lattice Curriculum.. 5. Implementation As mentioned in previous sections, based upon concept of OOP and SCORM standard, we have proposed an Instruction Activity Model (IAM) which is composed of related AT components with inter-relation and specific attributes for pedagogical needs. In this section, we want to verify that proposed IAM is workable and compatible with SCORM standard, so we develop an IAM system which is based on SCORM 1.3 RTE [2]. As shown in Figure 10, the IAM system consis ts of 2 modules: 1. IAM module including IAM Controller, IAM Model, and IAM database, and 2.. With vigorous development of the internet, e-learning system has become more and more popular. Currently, to solve the issue of sharing and reusing of teaching materials in different e-learning system, several international standard including SCORM, IMS, LOM, and AICC, etc. have been proposed by international organizations. Among international standards, Sharable Content Object Reference Model (SCORM) has become the most popular standard. In the Simple Sequence Specification (SSS) of SCORM, which relies on the concept of learning activities, defines the course sequencing behavior and organizes the content into a hierarchical structure, namely Activity Tree (AT). The Simple Sequencing process uses information about the desired sequencing behavior to control the sequencing, selecting and delivering of activities to the learner. Hence, how.
(10) to create, represent and maintain the Activity Tree and associated sequencing are our concerns. For a large learning activity, the Activity Tree will become too co mplex to manage and reuse. It is hard to reuse and reintegrate ATs without the inter-relations. These defects will limit the scalability and flexibility of adaptive learning system. Moreover, for more personalized learning, many researches have used the Pedagogical Theory to enhance the evaluation of the personal learning characteristic.. [10] Bengamin S. Bloom, Bertram B. Mesia, and David. Therefore, in this paper, we extend and modualize the structure of AT by applying Pedagogical Theory and concept of Object Oriented Methodology, respectively. Thus, we propose a novel model as graph, Instructional Activity Model (IAM), which is composed of related Activity Tree nodes. Based upon Pedagogical Theory, each AT node in IAM is modualized as a learning unit with inter-relations and specific attributes, which can be easily managed, reused, and integrated. We also propose two heuristic algorithms with general and specific domain heuristic to traverse IAM to generate the dynamic learning content to the learner. IAM with scalability and flexibility can apply different pedagogical theories for specific need by extension mechanisms. In addition, based on SCORM 1.3 RTE (Run Time Environment), we develop an IAM system to manage AT nodes and dynamically generate personalized SOCRM compliant course. Based upon the results of experiment, we may conclude that the IAM system is workable and compatible with SCORM standard. In the near future, we will enhance continuously our proposed IAM and apply it to some specific domains to evaluate its flexibility, scalability and learning performance.. in Cognitive Learning Theory with Implications,” Studies in Art Education, 36(3), pp. 134-153, 1995.. 7. Reference S. S. Tseng, G. J. Hwang, M. F. Jiang, C. J. Tsai, and Y.T. Lin, 專家系統 導論/工具/應用, 文魁 資訊股份有限公司, Taiwan, 2002. [2] SCORM (Sharable Content Object Reference Model), http://www.adlnet.org/ [1]. [3]. IMS (Instructional Management http://www.imsproject.org/. System),. [4]. LTSC (IEEE Learning Technology Standards Committee), http://ltsc.ieee.org/wg12/. [5]. AICC (Aviation Industry CBT Consortium), http://www.aicc.org. [6]. XML (eXtensible Markup http://www.w3c.org/xml/. [7]. W3C, World Wide http://www.w3.org. [8]. Edward R. Jones, SCORM, http://www.scorm.tamucc.edu/scorm/home.html. [9]. Nelson L. Bassing, Teaching in Secondary Schools, Boston: Houghton Mifflin Company, pp. 51-55. 1963.. Web. Language), Consortium,. R. Krathwohl, Taxonomy of Educational Objectives (two vols: The Affective Domain & The Cognitive Domain), New York. David McKay, 1964. [11] J. S. Bruner, The Process of Education, Cam-. bridge, MA: Harvard University Press, 1960. [12] Vincenza Carchiolo, Alessandro Longheu, and. Michele Malgeri, ”Adaptive Formative Paths in a Web-based Learning Environment, ” Educational Technology & Society, 5(4), 2002. [13] A.D. Efland, “The Spiral and the Lattice: Changes. [14] R.M. Gagne, “Learning Outcomes and Their Ef-. fects,” American Psychologist, No.4, 1984. [15] R.M. Gagné, L. Briggs and W. Wager, Principles. of Instructional Design, 4th ed. Harcourt Brace Jovanovich, 1992. [16] H. Gardner, Frames of Mind, New York: Basic. Books Inc., 1983. [17] K. Papanikolaou, M. Grigoriadou, G.D. Magoulas,. and H. Kornilakis, “Towards New Forms of Knowledge Communication: the Adaptive Dimension of a Web-based Learning Environment, “ Computers and Education, vol. 39, pp. 333-360, 2002. [18] G. J. Posner, Analyzing the Curriculum, N.Y. :. McGraw-Hill, 1992. [19] Leonif Sheremetov and Adolfo Guzman Are-. nas, ”EVA: An Interactive Web-based Collaborative Learning Environment,” Journal of Computers and Education, February 2002. [20] Sherry Chen, " A Congitive Model for Non-Linear. Learning in Hypermedia Programmes," British Journal of Educational Technology, 33(4), pp.449-460, 2002. [21] R. Sternberg and E. Grigorenko, Styles of Thinking. in the School, Branco Weiss Institute, May 1998. [22] Evangelos Triantafillou, Andreas Pomportsis, and. Stavros Demetriadis, “The design and the formative evaluation of an adaptive educational system based on cognitive styles,” Computers and Education, 41(1), pp. 87-103, August, 2003. [23] C. J. Tsai, S. S. Tseng, and Y. C. Wu, ”A New. Architecture of Object-Oriented Rule Base Management Sy stem,” Proc. of Tools Asia’99, Nanjing, China, Sep., 1999. [24] C. J. Tsai, and S. S. Tseng (2002), “Building A. CAL expert system based upon two-phase knowledge acquisition, “ Expert Systems with Applications 22,pp. 235-248. [25] Y.C. Wu (1999), An Approach to Object-oriented. Rule Base Management System, master thesis, Department of Computer and Information Science, National Chiao Tung University..
(11)
相關文件
On the course content page, click the function module to switch to different learning activities pages for learning; you can also directly click the "learning activity" in
Robinson Crusoe is an Englishman from the 1) t_______ of York in the seventeenth century, the youngest son of a merchant of German origin. This trip is financially successful,
fostering independent application of reading strategies Strategy 7: Provide opportunities for students to track, reflect on, and share their learning progress (destination). •
Students are asked to collect information (including materials from books, pamphlet from Environmental Protection Department...etc.) of the possible effects of pollution on our
volume suppressed mass: (TeV) 2 /M P ∼ 10 −4 eV → mm range can be experimentally tested for any number of extra dimensions - Light U(1) gauge bosons: no derivative couplings. =>
• Formation of massive primordial stars as origin of objects in the early universe. • Supernova explosions might be visible to the most
• elearning pilot scheme (Four True Light Schools): WIFI construction, iPad procurement, elearning school visit and teacher training, English starts the elearning lesson.. 2012 •
(Another example of close harmony is the four-bar unaccompanied vocal introduction to “Paperback Writer”, a somewhat later Beatles song.) Overall, Lennon’s and McCartney’s
