View
1
Download
0
Category
Preview:
Citation preview
Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006 1
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
ABSTRACT
Multimedia data and application systems accessible over the Web are valuable assets for developinginstructional materials for teaching, training, problem solving, and decision support. These assetscan be used to construct learning objects, each of which is a reusable granule of instructiondesigned to meet a specific instructional objective. In order to find and use learning objects, aninfrastructure for the registration, discovery, binding, and invocation of these objects is needed.Also, there is a need for an e-learning service infrastructure for people in a virtual e-learningcommunity to construct, evaluate, and deliver learning objects. In this work, we model distributedand sharable learning resources by two types of Learning Objects (LOs): Atomic Learning Objectand Composite Learning Object. Both types of LOs are published uniformly as Web services in aconstraint-based broker in order to make them sharable and reusable. This article presents thelearning object models for the specifications of these two types of LOs and an e-learning serviceinfrastructure, which consists of authoring tools for constructing LOs, software components forprocessing LOs and performing assessments, and an extended Web services framework for theregistration, discovery, binding, and invocation of LOs as Web services. This article also presentstechniques such as dynamic binding of LOs, rule-based execution of learning processes, and model-based assessment used to make the processing of LOs active, flexible, customizable, and adaptive.The roles and functions of virtual e-learning community members are also discussed.
Keywords: adaptive learning; database; e-learning; learning object; learning objectrepository; online learning community; virtual communities; Web-based learning;XML
Learning Object Models and anE-Learning Service Infrastructure
Gilliean Lee, Lander University, USA
Stanley Y. W. Su, University of Florida, USA
INTRODUCTIONMultimedia data and application sys-
tems that are accessible on the Web arevaluable assets for constructing instructionalmaterials for teaching, training, problemsolving, and decision support. One approachto make use of these distributed, heteroge-
neous data and application systems is toapply object-oriented technology and wrapthem as distributed objects. These distrib-uted objects then can be used to composelearning objects (LOs) for instruction andtraining purposes.
IDEA GROUP PUBLISHING
This chapter appears in the publication, International Journal of Distance Education Technologies Volume 4, Issue 1edited by Timothy K. Shih © 2006, Idea Group Inc.
701 E. Chocolate Avenue, Suite 200, Hershey PA 17033-1240, USATel: 717/533-8845; Fax 717/533-8661; URL-http://www.idea-group.com
ITJ3027
2 Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
In recent years, there has been a num-ber of initiatives in developing and standard-izing technologies for Web-based learning.The Advanced Distributed Learning Initia-tive (ADL, 2003), the IMS Global Learn-ing Consortium (IMS, 2004), and the OpenKnowledge Initiative (Eduworks andO.K.I. Leadership, 2002) are a few ex-amples. The Sharable Content Object Ref-erence Model (SCORM) (ADL, 2004a) isa reference model initiated by the Ad-vanced Distributed Learning (ADL) pro-gram of the Department of Defense (DoD)and the White House Office of Scienceand Technology Policy (OSTP). Accord-ing to the SCORM’s specification, it is en-visaged that Internet users and heteroge-neous LMSs would use the Web as a uni-versal platform for accessing and launch-ing sharable content objects and for estab-lishing close communication, interaction, andcoordination among content object devel-opers, course authors, users, and adminis-trators. To realize this vision, sharable con-tent objects must be durable, interoperable,accessible, and reusable. In order to meetthese requirements, it is necessary to havea uniform way of modeling not only learn-ing resources but also heterogeneous learn-ing tools and LMSs, as well as an informa-tion infrastructure to enable theinteroperation and sharing of their contentsand functionalities. Also, the aggregationmodel that defines the learning sequenceor process has to be flexible, adaptable, andcustomizable in order to meet differentlearners’ needs and learning contexts. Ourresearch and development work to meetthe above needs is consistent with the vi-sion and goals of the ADL Program.
A virtual community is a Web sitethrough which members of the communitycan share useful information related to theircommon interests. Virtual educational com-munities, such as MERLOT (2004), EOE
(2004), and CLOE (2004), provide learn-ing object repositories, from and to whichpeople can find and provide useful learningmaterials. They provide Web-browser-based user interfaces for the registrationand search of learning objects. However,they do not support application interfacesfor programmatic searches and accessesto learning materials. Such interfaces wouldfacilitate the reuse of learning objects bylearning management systems that bindlearning objects dynamically at runtime.
In York et al. (2002, p. 4), it was en-visioned as “building the technological in-frastructure to support dynamic, ad-hoccommunities of lifelong learners who in-teract within an environment of learningobjects through a creative blend of ad-vanced computing technologies, high per-formance networks, authoring and collabo-ration tools.” In the same context, it is use-ful to establish an infrastructure over theInternet to allow people who are interestedin specific subjects of learning to form theirown virtual communities.
In this work, we model distributed andsharable learning resources by two typesof Learning Objects (LOs): Atomic Learn-ing Object and Composite Learning Ob-ject. LOs are uniformly published as Webservices in a constraint-based, Web-serviceregistry and are made sharable and reus-able. This article presents the learning ob-ject models for the specifications of thesetwo types of LOs and an e-learning ser-vice infrastructure, which consists ofauthoring tools for constructing Los; soft-ware components for processing LOs andperforming assessments; and an extendedWeb services framework for the registra-tion, discovery, binding, and invocation ofLOs as Web services.
The organization of this article is asfollows. Related works are covered in thenext section, which is followed by the pre-
Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006 3
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
sentation of the two LO models. The e-learning service infrastructure and its toolsand software components are then de-scribed. A summary is given in the finalsection.
RELATED WORKSEffective delivery of learning contents
to learners of different backgrounds andinterests in order to maximize the effect oflearning is one of the goals of e-learning.Structuring learning contents is related di-rectly to this goal. Learning process mod-els, such as SCORM’s Content Aggrega-tion Model (ADL, 2004b) and Cisco’s Re-usable Learning Object (Termaat et al.,2003) are two popular models that orga-nize learning contents/objects for effectivedelivery. A tree structure, used in SCORM,Cisco’s reusable learning object (RLO), L3(Altenhofen, 2002), and KnowledgeTree(Brusilovsky & Nijhavan, 2002), is gener-ally used to model a learning process. Atree node represents a granule of learningobject/content to be presented to learners,which can be a course, module, or lesson.
SCORM is intended to provide a stan-dard framework for building learning sys-tems that enable the reusability and shar-ing of learning contents developed by dif-ferent authoring tools. SCORM’s Sequenc-ing Definition Model (ADL, 2004a) is a rule-based sequencing model that supportsadaptive execution of an Activity Tree us-ing sequencing rules, rollup rules, and thestatus of learning objectives. The ActivityTree of the model is a hierarchical struc-ture of learning activities and their corre-sponding learning contents.
SCORM’s Sequencing DefinitionModel provides the desirable set of fea-tures previously described; however, it hasa few limitations. First, it does not have aconceptual model for modeling learningobjects. In SCORM, three types of learn-
ing resources are distinguished: Asset, Shar-able Content Asset (SCA), and SharableContent Object (SCO). However, whatthese resources are comprised of has notbeen specified explicitly. Second, non-leafactivities in an Activity Tree neither presentcontents nor perform assessment inSCORM. We believe that it is useful to al-low a non-leaf node to present an intro-duction and/or a summary of the contentscovered by its child activities. Assessmentitems for testing the integrated knowledgeof the contents presented by the child ac-tivities should also be allowed. Third, thereis no model or specification for assessmentsin the sequencing model; hence, the cover-age of an assessment is decided by theSCAs or SCOs bound to the activity. Achange of assessment coverage in an ac-tivity tree requires a structural change ofthe learning process and the authoring ofan SCO that conducts the new assessment.This may require a substantial amount oftime and effort. In our work, we introducea model-based assessment technique tofacilitate flexible assessments.
A workflow process model repre-sented as a directed acyclic graph (DAG)is employed in Flex-EL (Lin, Ho, Sadiq, &Orlowska, 2002). The model is used tospecify the interaction, coordination, andcollaboration among the people involved inlearning as well as the sequencing of con-tents to be presented to learners. This sys-tem uses conditional transitions expressedin a workflow model to achieve the flex-ibility in selecting the learning pathways thatmatch learners’ individual paces and learn-ing styles. Different from Flex-EL, a learn-ing process in our work is modeled by anactivity tree, and sequencing rules and rolluprules are used as in SCORM. The actionpart of the rules can change the sequenc-ing paths by allowing learners to retry, hide,disable, and skip activities when the condi-
4 Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
tions of the rules are met. Additionally,customization of the structure of a learningprocess or sequencing control modes atruntime is achieved by invoking rules tocheck a learner’s background and progressand to perform proper actions.
KnowledgeTree (Brusilovsky &Nijhavan, 2002) separates learning materi-als into primary materials for average learn-ers and additional materials for learners withdifferent learning styles and knowledge.The system uses learning goals, prefer-ences, and knowledge of the individuallearner to select the most appropriate learn-ing materials. Our dynamic binding of re-quests to LOs is similar to their idea of bind-ing learning materials at runtime. However,we leverage standardized Web-servicestechnologies for the registration and dis-covery of LOs and add metadata and con-straints in LO specifications.
L3 (Altenhofen, 2002) separateslearning processes from strategies (i.e.,navigation rules) and introduces meta-tagsfor describing the knowledge type of learn-ing activities and the interrelationshipsamong the activities. A strategy chosen atruntime and the meta-tags of learning ac-tivities enable adaptive navigation of a learn-ing process. Different from their work, ourlearning process model allows sequencingrules and rollup rules to be attached to eachactivity at both design time and runtime inorder to achieve dynamic properties in alearning process execution.
LEARNING OBJECT MODELSInstructional materials need to be de-
veloped in order to deliver learning experi-ences to learners. The cost of developingthese materials can be amortized, if theyare used repeatedly. The object conceptintroduced in object-oriented programmingcan be applied to encapsulate learning con-tents as objects for reuse.
Definition of Learning ObjectA learning object is a granule of in-
struction designed to meet a specific in-structional objective(s). We model a learn-ing object in terms of content items, prac-tice items, and assessment items neededto cover a subject of learning and to con-duct an assessment. By including thesethree types of learning items in an LO speci-fication, the author of the LO can ensurethat practice and assessment items pre-sented to learners are consistent with thepresented content items. In addition tothese three types of items, the metadata,which is a description of the LO, can beadded in an LO specification (Johnson,2003; Termaat et al., 2003). After an LO isauthored, it is published as a Web servicein an extended Web services infrastructure,which will be discussed in a later section.
The metadata is expressed in termsof a set of attributes and the constraintsassociated with these attributes. The at-tributes include Title, Subject, Author, Key-words, Target Age, Media Format, Lan-guage, Cost, Quality Rating, and so forth,as in MERLOT, CLOE, and IEEE LOMstandard (IEEE LTSC, 2002). LOs targetedfor a specific domain may require additionalmetadata attributes. In order to satisfy thisrequirement, we allow the insertion anddeletion of metadata attributes. These at-tributes can be categorized into two types.Static attributes such as Title, Subject, Key-words, Language, Target Age; and MediaType are specified by the authors. Evalua-tors (a role played by some qualified peoplein a virtual community) as well as the LOrepository and the LO execution systemrecord and update dynamic attributes suchas Usage Number, Average Score, andQuality Rating. Constraints can be catego-rized into attribute constraints (i.e., the le-gitimate values of these attributes) andinterattribute constraints (i.e., their value
Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006 5
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
relationships). An attribute constraint hasthe following form:
• attributeId keyword value | range |enumeration,where attributeId is the identifier of anattribute, keyword can be either a com-parison operator “=”, “!=”, “<”, “<=”,“>”, or “>=”, or a “RANGE” or “ENU-MERATION” specification. In theformer case, value specifies the valuethat the attribute must have. In the lat-ter case, range denoted by [value..value]specifies the value range that the at-tribute can have, or enumeration de-noted by {value, value, .., value} speci-fies a list of possible values of the at-tribute.
An interattribute constraint is definedas follows:
• If attribute_constraints then attribute_constraints,where attribute_constraints is a Bool-ean expression in which each term is anattribute constraint.
For example, an attribute constraintof an LO specified by “Target_age = [3..8]”may target the LO to children betweenthree and eight years old. An LO that isfree of charge to people from a developingcountry can be specified by an interattributeconstraint, “If learner.country =developing_country, then cost = 0.” Themetadata is used for the registration, dis-covery, and binding of learning objects as aWeb service (Lee, Zhang, & Su, 2004).
Hierarchy of Learning ResourcesWe distinguish three types of learn-
ing resources and define a hierarchy basedon the order of their constructions. The low-
est level of resources in the hierarchy iscalled learning asset, which represents themost basic resources available on the Web.A learning asset can be a simple file suchas a text, image, audio or video clip, plug-infiles such as Flash, or a complicated Webpage. The next level of resources is calledAtomic Learning Object (ALO), whichrepresents a basic unit of instruction thatcontains content items for teaching learn-ers some problem-solving skill or providingthem the knowledge needed in some deci-sion-making situations. An ALO also cancontain practice and assessment items forexercising and assessing learners’ acquiredskills or knowledge. An ALO is specifiedin XML-based Learning Content Defini-tion Language (Su & Lee, 2004).
ALOs by themselves will have lim-ited use, unless they can be used to com-pose larger granules of instruction. A Com-posite Learning Object (CLO) is an aggre-gation model defined as a tree structure ofactivities with sequencing information. Inthe activity tree, learning activities arenodes, connectors are used to bridge par-ent activities and their child activities withthe sequencing control mode (ADL,2004a) that defines the sequencing behav-ior among its child nodes, and edges areused to connect activities to connectors, andvice versa. By separating the control in-formation (sequencing control modes) fromthe activity specification, runtimecustomization of a learning process instancecan be carried out more easily. Figure 1shows a CLO on Relational Model, com-posed of the modeling constructs discussedpreviously.
Learning objectives and activity datacan be defined in each activity. A primarylearning objective, which is required foreach activity, represents the activity’s sat-isfaction status. A learning objective has
6 Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
build-time data fields such as Objective ID,Satisfied by Measure, Minimum SatisfiedMeasure, and Contributes to Rollup. Atruntime, a learning objective will have aBoolean value for its satisfaction status anda numerical value for its normalized score.The details of learning objective descrip-tions can be found in the SCORM Sequenc-ing Definition Model (ADL, 2004a). Ac-tivity data are used for internal processingof learner profiles, objectives, andmetadata.
Non-leaf activities can have content,practice, and assessment items. This is use-ful for providing an abstract and/or intro-duction to learners before the subtreerooted at a non-leaf activity is launched anda summary of the content items of thesubtree after it is processed. Assessmentand practice items in a non-leaf activity canbe designed to assess learners’ abilities tointegrate the concepts and contents pre-sented in the tree rooted at the activity. Leafactivities can bind to specific ALOs and
CLOs statically or can contain requests fordynamic binding to suitable ALOs andCLOs. Additionally, a set of rules can beincluded in a CLO specification and be usedto alter the navigation sequence and thestructure of an activity tree based on indi-vidual learners’ profiles (i.e., backgrounds,ages, preferences, and competencies) andprogress. This rule-based execution allowsthe CLO to be processed in a flexible, adapt-able, and customizable fashion. An XML-based Learning Content Definition Lan-guage (Su & Lee, 2004) has been devel-oped to describe CLOs, including their se-quence information and rules. Figure 2 il-lustrates the taxonomy and reference/bind-ing relationships among learning resources.
Dynamic E-Learning Service:Requirements and Techniques
Internet users have much more di-verse backgrounds than students. There-fore, Web-based learning has to be dynamicin order to accommodate learners’ differ-
Figure 1. Modeling constructs in a composite learning object
Edge
Non-leaf Activity
Leaf Activity Connector
Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006 7
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
ent backgrounds, competencies, and inter-ests. To meet this requirement, learningobject service must have the following dy-namic properties: active, flexible, adaptive,and customizable. They should be activeso that operations on them automaticallycan trigger rules to enforce policies andconstraints, coordinate learning activities,and notify peers and content producers toachieve collaborative e-learning. Theyshould be flexible so that small granulesof LOs and multimedia content objectscan be used to flexibly compose largerLOs, and a request for an LO can be flex-ibly and dynamically bound to a suitableLO. LOs should be adaptive so that theway content, practice, and assessmentitems are selected and presented to learn-ers can adapt to suit individuals’ profilesand needs. They should be customizableso that their learning process models (i.e.,the activity tree structures and sequencingcontrol modes) can be customized atruntime based on the progress and profile
of learners. The following sections describethe techniques and mechanisms used tosatisfy the requirements.
Rule-Based Execution ModelEach activity specification contains an
activity name, an activity ID, a textual de-scription of the activity, learning objectives,activity data, a set of optional condition-action rules, optional assessment informa-tion for assessment execution and problemselections, and a limit on the execution ofactivity in terms of time or the number ofallowed executions. It also can containlearning items in a non-leaf activity or bind-ing information in a leaf activity, which mayinclude service request information such asa business service key for static binding orconstraints for dynamic binding. As shownin Figure 3(a), the condition-action rules ofa non-leaf activity may contain (1) pre-ac-tivity rules, (2) after-pre-assessment rules,(3) drill-down rules, (4) rollup rules, and (5)after-post-assessment rules. Before an
Figure 2. Hierarchical relationship of learning resources
CLO Meta-data, Constraints
Leaf Activity
Leaf Activity
Leaf Activity
References
Binds to
Non-leaf Activity
Intro, summary
Practice items
Assessment items
Plug-in files
Video clip
Image
Web pages
Learning Assets
Audio clip Meta-data & Constraints
ALO Intro, summary
Practice items
Assessment items
CLOs
8 Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
activity is executed, after a pre-assessmentis conducted, and before navigating downto a child activity, relevant rules are ex-ecuted for any possible adaptation orcustomization. When a rollup process isinitiated upon the completion of a child ac-tivity, its parent would post a rollup eventto trigger any applicable rollup rule. If thereare one or more child activities yet to beprocessed, the sequencing control mode willdetermine the next child activity that shouldbe processed. In that case, a drill-down
process to that child activity will take place.Otherwise, a summary followed by the pre-sentation of practice and assessment itemsare delivered to the learner. Before com-pleting a non-leaf activity, the after-post-assessment rule is checked and executed.The significant differences between leafand non-leaf activities are that drill-downand rollup rules do not exist in a leaf activ-ity, and an LO is bound to a leaf activityafter the pre-activity event is posted, asshown in Figure 3(b).
Figure 3. Tasks in an activity node at runtime: (a) non-leaf activity (top) and (b)leaf activity (bottom)
After-Post-assessment Event
Post-assessment (if specified)
Present Practice Items
Present Content item (Summary)
Roll-up Event
Pre-activity Event
Pre-assessment (if specified)
After-Pre-assessment Event
Present Content items (Overview)
Drill-down Event
Roll-up Notified
Drill-down Notified
Event Posting Execution Flow
Condition- Action Rules
Condition- Action Rules
Notify Drill-down
Notify Roll-up
Notify Roll-up
Pre-activity Event
Bind to a Learning Object
Pre-assessment (if specified)
After-Pre-assessment Event
Present Content items
Present Practice Items
Post-assessment (if specified)
After-Post-Assessment Event
Condition- Action Rules
Drill-down Notified
Yes
No
Child activity to visit?
Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006 9
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
At each of the above five stages ofprocessing an activity, an event is postedto trigger the processing of condition-ac-tion rules that are associated with the event.Since the condition part of these rules maycheck learners’ profiles, assessment results,and the progresses that they have made atthe various stages of processing the activ-ity, different actions may be taken by rulesto alter the learning paths for learners withdifferent competencies and profiles. Trig-gered rules also can customize the controlstructure of a CLO runtime instance in or-der to alter its control flow, activity specifi-cation, and sequencing control modes to ac-commodate learners of distinctively differ-ent profiles. To facilitate adaptive and flex-ible LO binding, the binding information ina leaf activity can reflect the context andconstraints (time, language, cost, age, etc.)dynamically, as well. The action part of arule may specify Skip Activity, DisableActivity, Hide Activity from Choice, ExitActivity, Retry Activity, Assess, Update thesequencing control mode of a Connector,Replace Activity (A1, A2), and so forth, sothat the delivery sequence of activities canbecome adaptive, flexible, andcustomizable. For example, the pre-activ-ity rule, If activity (A1) is satisfied, thenHide activity (this) From Choice, speci-fied in activity A2 would prevent a learnerfrom choosing activity A2 when the learnerhas successfully completed A1. An after-pre-assessment rule, If pre-assessmentobjective is satisfied, then Skip activity(this), would allow the current activity tobe skipped when the pre-assessment re-sult is satisfactory. A drill-down rule of anon-leaf activity specified as Iflearner.job_type != ‘technician’, thenSuppress subtree would prohibit a learnerfrom navigating through the subtree thatcovers technical details when the learneris not a technician. Such rules can enable a
CLO to provide learners different contents,depending on the learners’ profiles andprogresses.
Dynamic BindingLO repositories established at the
sites of those authors who created LOs maybe updated frequently (i.e., new LOs arecreated, and existing LOs are enhancedwith up-to-date contents or removed fromthe repositories). One way to leverage theLOs stored in these repositories is to regis-ter the meta-information and access infor-mation of these distributed LOs with an LObroker. The users of LOs (e.g., learnersand content authors) can then query thebroker to find the desired LOs. Also, a soft-ware component can have programmaticaccesses to the access information associ-ated with the registered LOs by queryingthe broker at runtime. We call the runtimebinding of a request to an LO a dynamicbinding.
Dynamic binding of LOs is supportedby allowing a leaf-activity to specify a queryat build-time, and to issue the query to theLO Broker at runtime. The query describesthe requirements of an LO that can satisfythe learning objective of a leaf-activity. Thequery requirements are specified in termsof attribute and interattribute constraints;the same as the metadata specifications ofregistered LOs. The broker would matchthe query requirements against the meta-information of registered LOs to find thesuitable LO(s). If more than one LO thatmatched with the query’s requirements arereturned, the metadata of these LOs willbe shown to the learner so that he or shecan choose an LO that suits his or her need.
A query may contain a variable in itsattribute constraint specification. This vari-able will be assigned a value based on alearner’s profile, status of objectives, andactivity data. In a query specification of a
10 Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
leaf activity of a CLO, the value on therighthand side of an attribute constraint canbe prefixed by LP:, OBJ:, or VAR:, whichrepresents the learner profile, objective, oractivity data, respectively. At runtime, thevariable with the prefix will be assigned avalue. For example, an attribute constraint,Target_age = LP:age, will be transformedinto Target_age = 15 when a 15-year-oldlearner is taking the CLO. As another ex-ample, a learner with a disability can be pro-vided with appropriate content by an attributeconstraint, Media_type ENUMERATION{VAR:content_type}, where the activity datacontent_type can be replaced by an enu-meration of strings: {Movie with caption,HTML, or Flash} for a hearing-impairedlearner or {Audio, or Movie with audio} fora visually impaired learner.
Model-Based AssessmentIn SCORM, only leaf nodes of an
activity can have assessments. In our work,any activity node in a learning processmodel can carry out a pre- and/or post-assessment task on the learning contentscovered by the tree that is rooted at thenode. The assessment items used in theassessment task can be selected randomlyand proportionally from the assessmentitems of the non-leaf nodes and those learn-ing objects bound to the leaf nodes of thetree. The previous provision gives a con-tent producer greater flexibility specifying
when and how assessments are to be car-ried out. Figure 4 shows an activity treewith several assessments. A set of assess-ment questions will be given after each ofthe activities in level 2 under the activityA2, and only those learners who pass lessthan two of the three assessments need totake the assessment on activity A2. Tomake the assessment on A2 optional, arollup rule of activity A2 can be defined asif At Least 2 child activities are Satis-fied, then mark activity(this) as Satisfied,and an after-post-assessment rule ifactivity(this) is not Satisfied, then per-form the Assessment. In the child activi-ties of A3, there will be no assessment, andthere will be a mandatory comprehensiveassessment for the entire tree rooted at A1.The rollup rule of the non-leaf activity A3is If All child activities are Satisfied, thenmark activity(this) as Satisfied. Activi-ties without any assessment will be con-sidered as Satisfied after they have beennavigated. The root activity A1 has an ob-jective assessment_objective for the post-assessment conducted at the root activity.The entire CLO will be considered as satis-fied, only if activity A2 and the objectiveassessment_objective are satisfied, whichcan be specified by an after-post-assessmentrule of A1: if (activity(A2) is Satisfied) AND(objective(assessment_objective) is Satis-fied), then mark activity(this) as Satisfied.
Figure 4. An activity tree with flexible assessments
Activity without assessment
Activity with assessment
Activity with optional assessment
Connector
Level 0
Level 1
Level 2
Level 3
A6
A2
A1
A3
A5 A4
Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006 11
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
E-LEARNING SERVICEINFRASTRUCTURE FORVIRTUAL COMMUNITIES
The existing virtual e-learning com-munities provide directory services for theregistration and discovery of learning ob-jects and spaces for exchanging informa-tion among members. However, they donot provide learning management servicessuch as assessment, adaptive learning pro-cess enactment, and learner profile man-agement, which are essential for achievingeffective and flexible e-learning.
To provide these services mentioned,an infrastructure is needed to allow the reg-istration, discovery, binding, and invocationof LOs, as well as the sharing of multime-dia data objects and Web-based applica-tion systems. It also should allow the es-tablishment of multiple e-learning virtualcommunities. The Web-service technologybeing developed by the IT industry can beadapted and extended to serve this purpose.Universal Description, Discovery, and In-tegration (UDDI) (OASIS, 2004) providesthe general framework to allow all typesof objects to be defined as Web servicesusing the Web Service Description Lan-
guage (WSDL) (W3C, 2001). However,the existing specification and implementa-tion of UDDI does not allow the specifica-tion of metadata and constraints associatedwith Web services. Without this specifica-tion capability, a learning object can be se-lected by the registry for downloading orfor online use but does not satisfy arequestor’s requirements.
In this work, we have extended theWSDL and UDDI capabilities by allowingthe specification and processing of con-straints associated with learning objects(Lee, Zhang & Su, 2004). Requestors’ re-quirements also can be specified in a con-straint specification, as described in the lastsection. We use a Constraint SatisfactionProcessor (Degwekar, Su & Lam, 2004)that was developed at the Database Sys-tems R&D Center to store constraints andperform constraint matching to select suit-able Web services.
A CLO is processed by the LearningProcess Management System, which con-tains a Learning Process Execution Engine(LPEE), an Event-Trigger-Rule (ETR)Server (Lee, Su & Lam, 2001), a User In-terface Component, an Assessment Com-
Figure 5. Architecture of learning process management system
Web-S
ervice (Web, S
OA
P server)
Learning Process Management System
Assessment Component
Learning Process Execution Engine (LPEE)
UI Component
ETR Server
LO Repository, Runtime Data,
Learner Profiles
- LO Repositories - Broker service
Web Client
12 Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
ponent, and an LO Repository as its keycomponents (see Figure 5). From theLPEE’s point of view, an ALO is a specialcase of a CLO having a single leaf activitythat binds to the ALO. When an ALO isrequested by a learner for processing, theALO is transformed into a CLO and thenexecuted by the LPEE. The ETR Servermanages and processes events, triggers,and rules. The local LO repository is usedto maintain the data required for the pro-cessing of CLO instances, such as down-loaded LOs, learner profiles, and runtimedata models. The Assessment Componentgathers assessment items to present tolearners, collects their answers for grad-ing, and stores the assessment results inthe LO Repository. The UI componentgenerates Web pages that deliver LOs tolearners and facilitate communication be-tween learners and the system. The com-munication between the Learning ProcessManagement System and the external com-ponents, such as the remote LO Reposito-ries and the Broker, are facilitated by Webservices using Apache Tomcat and the AxisSOAP engine.
Virtual E-Learning Community:Roles and Functions
People or organizations play differ-ent roles in a virtual e-learning community.Content providers use authoring tools todefine and create learning assets andALOs. The ALOs are stored and main-tained at the providers’ local repositoriesthat make use of the Apache Xindice XMLdatabase (Apache Software Foundation,2004). Content composers use a CLOauthoring tool (Lee, Zhang & Su, 2004) tocompose CLOs. Metadata and accesspoints of ALOs and CLOs are registeredas Web services in a Constraint-BasedBroker, which can be browsed and que-ried by users who play various roles. The
broker leverages the UDDI registry as aback end to store the access points of LOs.The operations (getLO, getConstraints,etc.) of these objects are accessible asWeb-service operations by the Simple Ob-ject Access Protocol (SOAP) (W3C,2003). Each virtual community has a Com-munity Host site that installs the LearningProcess Management System and the Con-straint-Based Broker. The main functionsof Community Host Administrator are toadminister the activities of the virtual com-munity, maintain the Learning ProcessManagement System, and enforce thecommunity’s rules and policies. ContentEvaluators, elected by the virtual commu-nity, review, evaluate, and rate the ALOsand CLOs that are registered with the reg-istry. They approve or disapprove their in-clusion for the community’s use as well astheir modification and removal. A ContentLearner registers himself or herself in avirtual community with his or her profile(i.e., background, preference, accessibility,etc.) for the dynamic e-learning service. Alearner’s profile, described using the learnerprofile model, is stored in the LO reposi-tory of the Community Host. The roles ofthe people in a virtual e-learning commu-nity and their functions related to LOs areshown in Figure 6.
A Runtime Scenarioand Infrastructure
A registered content learner woulduse the Web page generated by the userinterface component to browse a direc-tory of registered LOs maintained by theConstraint-Based Broker of the Commu-nity Host. He or she would specify his orher learning requirements in terms ofmetadata and constraints associated witha desired LO. Those LOs that match withthe Content Learner’s requirements willbe selected by the broker, and their
Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006 13
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
metadata are presented to the learner. Thelearner then can select the one he or shewants, and the business service key of theselected LO will be passed to the Learn-ing Process Execution Engine (LPEE),which will contact the broker to get anLO ID and an access point to the LO re-pository, from which the selected LO canbe downloaded to the LO repository ofLPEE. After downloading, an LO instanceis deployed for execution. LPEE wouldcreate a runtime data model in the localrepository and generate code for activi-ties, events, rules, and triggers for rule-based execution.
During the execution of an LO in-stance, a leaf activity binds to an ALO orCLO. In the case of static binding, a spe-cific LO is bound to the leaf activity. Butin dynamic binding, a query is issued to
the LO broker to obtain the most suitableLO for the learner. After an LO is cho-sen, LPEE again downloads and deploysthe LO for execution. The runtime statusof the LO instance is stored in the localrepository.
When the execution of an LO instanceis finished, metadata attributes such as us-age number, average score, and the learner’sgrading of the LO are updated and stored inthe Constraint-Based Broker. Figure 7 showsthe e-learning service infrastructure of a vir-tual e-learning community and the opera-tions and communications among its soft-ware components.
SUMMARY AND CONCLUSIONIn this article, we have presented
problems and requirements in the designand implementation of LOs. We defined
Figure 6. Roles in a virtual e-learning community and their functions
Content Composers
Content Evaluators
Content Learners
ALOs Asset
CLOs
Provides E-learning service to
Generate
Generate
Used by
Evaluated by
Registered to,
Used by
Host Admin
Evaluated by
Registered to,
Used by
Evaluation Tool
Community Host
Community Rules
Maintain LO registration
Administers
Uses
Log evaluation on LOs
Generate
Used by
Content Providers
Content Providers
Establishes Register Profile
14 Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
two types of LOs: ALOs that contain con-tent, assessment, and practice items; andCLOs that are composed of ALOs andCLOs. The metadata of both types of LOsare registered with the Constraint-BasedBroker for their discovery and binding. Theconcept of virtual e-learning community andthe collaborative roles of its members havebeen explained. A Web-service-based e-learning service infrastructure and the keycomponents of a Learning Process Man-agement System also have been described.The Learning Process Management Sys-tem makes use of several components thatwere developed for other applications, suchas the ETR Server and the Constraint-Based Broker. Its authoring tools for de-signing ALOs and CLOs, distributed LOrepositories, the Learning Process Execu-tion Engine also have been developed. Inaddition to system development, researchproblems such as security and privacy is-sues related to e-learning, methodologiesfor the design and evaluation of e-learning
objects, and the problem of resolving dif-ferent ontologies used by authors and us-ers to define and query for learning objectsare being investigated.
REFERENCESAdvanced Distributed Learning (ADL).
(2003). Advanced distributed learn-ing. Retrieved June 16, 2004, from http://ADLNet.org
Advanced Distributed Learning (ADL).(2004a). Sharable content object ref-erence model (SCORM) 1.3: Se-quencing and navigation. RetrievedJanuary 30, 2004, from http://ADLNet.org
Advanced Distributed Learning (ADL).(2004b). Sharable content object ref-erence model (SCORM): The SCORMcontent aggregation model. RetrievedJanuary 30, 2004, from http://ADLNet.org
Altenhofen, M., & Schaper, J. (2002). Flex-ible instructional strategies for e-learn-
Figure 7. Web-service-based e-learning service infrastructure
Link Assets (URL)
Browse ALOs Register CLOs. Lookup CLOs, ALOs
Request ALOs Request CLOs
Content Composer
CLOs
Create, Edit CLOs
Web-page authoring tool
Import
Link assets (URL)
Create Edit ALOs
Content Provider
ALOs
HTMLs
Broker
UDDI Registry
ALOs CLOs
Community Host
Register ALOs
Learning Process
Management System
CLO Authoring Tool
Web Servers on Internet
Learning Assets Content Learner
Request, Provide CLO Execution, Content Delivery service
Register, Lookup LOs
Web, SOAP Server
Learning Assets
ALO Authoring Tool
Web, SOAP Server
Web, SOAP Server
Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006 15
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
ing. In Proceedings of the 35th Ha-waii International Conference on Sys-tem Sciences, Big Island, Hawaii.
The Apache Software Foundation. (2004).Apache xindice. Retrieved June 17,2004, from http://xml.apache.org/xindice/
Brusilovsky, P., & Nijhavan, H. (2002). Aframework for adaptive e-learning basedon distributed re-usable learning activi-ties. In Proceedings of World Confer-ence on E-Learning, E-Learn 2002(pp. 154-161), Montreal, Canada.
Co-operative Learning Object Exchange(CLOE). (2004). CLOE: Co-operativelearning object exchange. RetrievedJune 16, 2004, from http://cloe.on.ca/
Degwekar, S., Su, S. Y. W., & Lam, H.(2004). Constraint specification and pro-cessing in Web services publication anddiscovery. In Proceedings of the In-ternational Conference on Web Ser-vices, San Diego, California.
Eduworks, & O. K. I. Leadership (2002).What is the open knowledge initia-tive? (OKI white paper). Retrieved June15, 2004, from http://web.mit.edu/oki/product/whtpapers/whatis.html
The EOE Foundation. (2004). RetrievedJune 17, 2004, from http://www.eoe.org/index.htm
IEEE Learning Technology Standards Com-mittee (LTSC). (2002). Draft standardfor learning object metadata. Re-trieved October 7, 2004, from http://l t s c . i e e e . o r g / w g 1 2 / f i l e s /LOM_1484_12_1_v1_Final_Draft.pdf
IMS Global Learning Consortium (IMS).(2001). IMS learner information pack-age specification. Retrieved June 16,2004, from http://www.imsglobal.org/profiles/index.cfm
IMS Global Learning Consortium (IMS).(2004). Retrieved June 16, 2004, fromhttp://www.imsglobal.org
Lee, G, Zhang, X., & Su, S. Y. W. (2004).Learning object models, authoring tools,and learning object registry service forachieving dynamic and collaborative e-learning. In Proceedings of the 7thIASTED International Conference onComputers and Advanced Technologyin Education 2004 (pp. 321-327),Kauai, Hawaii.
Lee, M., Su, S. Y. W., & Lam, H. (2001).A Web-based knowledge network forsupporting emerging Internet applica-tions. The WWW Journal, 4(1/2), 121-140.
Lin, J., Ho, C., Sadiq, W., & Orlowska, M.E. (2002). Using workflow technologyto manage flexible e-learning services.Educational Technology & Society,5(4), 116-123.
MERLOT. (2004). Retrieved June 16,2004, from http://www.merlot.org/Home.po
OASIS. (2004). Universal description,discovery, and integration (UDDI).Retrieved June 16, 2004, from http://www.uddi.org/
OCW. (2003). MIT OpenCourseWare.Retrieved June 15, 2004, from http://ocw.mit.edu/index.html
Su, S. Y. W., & Lee, G. (2004, November1-5). A Web-services-based e-learninginfrastructure for managing virtual e-learning communities. In Proceedingsof E-Learn 2004: World Conferenceon E-Learning, Washington, DC.
Termaat, B., et al. (2003). Reusable learn-ing object strategy: Designing anddeveloping learning objects for mul-tiple learning approaches. RetrievedJune 29, 2004, from http://bus iness .c i sco .com/serv le twl3/F i leDownloader / iqprd /104108/104108_kbns.pdf
W3C. (2001). Web services descriptionlanguage (WSDL) 1.1. Retrieved June
16 Journal of Distance Education Technologies, 4(1), 1-16, January-March 2006
Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc.is prohibited.
Gilliean Lee was born in Seoul, Korea. He received his BS and MS in computerscience from Sogang University, Korea, in 1992 and 1994, respectively. He wasthen a senior software researcher at Hyundai Electronics Co., Hyundai InformationTech. Co., and Posdata Co. He is a recipient of a University of Florida AlumniFellowship. He received his MS and PhD in computer and information scienceand engineering at the University of Florida in 2003 and 2005, respectively. He iscurrently working as a faculty member in the Department of Math and Computingat Lander University. His research interests are in e-learning management, Webontology, e-business, and workflow management.
Stanley Y. W. Su earned his MS and PhD degrees in computer science from theUniversity of Wisconsin, Madison (1965 and 1968, respectively). He is distinguishedemeritus professor of the Department of Computer and Information Science andEngineering, University of Florida. He was one of the founding members of theIEEE Computer Society’s Technical Committee on Database Engineering. He haschaired eight major professional conferences in database and information systems.He served as an editor of IEEE’s Transactions of Software Engineering, an areaeditor of the Journal of Parallel and Distributed Computing, editor of IEEE’s Transactionson Knowledge and Data Engineering, editor of the Information Sciences journal,editor of the International Journal on Computer Languages, a member of the board oftrustees and treasurer of the executive committee of the Very Large Data BaseEndowment, and an editor-in-chief of the Very Large Data Base Journal. He currentlyserves as editor of the Asian Journal of Business & Information Systems, associateeditor of the International Journal of Web Services Research (JWSR), and a memberof the editorial advisory board of the International Journal on Computer Languages,Systems and Structures. He is an IEEE fellow.
10, 2004, from http://www.w3.org/TR/wsdl
W3C. (2003). Simple object access pro-tocol. Retrieved June 15, 2004, fromhttp://www.w3.org/TR/SOAP/, andhttp://www.develop.com/soap/
York, B., et al. (2002). A teacher for ev-ery learner. Retrieved April 8, 2004,from http://www.cra.org/Activities/grand.challenges/slides/education.pdf
Recommended