A WEB-SERVICE-BASED E-LEARNING SERVICE INFRASTRUCTURE FOR
ACHIEVING DYNAMIC AND COLLABORATIVE E-LEARNING
By
GILLIEAN LEE
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
2005
Copyright 2005
by
Gilliean Lee
To my wife, Rang-wook, and my parents.
ACKNOWLEDGMENTS
The work with this dissertation has been extensive, requiring much work and fun.
Without help and support from many persons, I would never have been able to finish this
work.
First of all, I would like to thank my advisor Stanley Su for his inspiring and
encouraging way to guide me, and his invaluable comments during the whole work with
this dissertation. He showed me how to approach a research problem and how to devise
and provide a systematic solution. He also provided me with many opportunities to
publish papers and meet people in the e-learning area.
I would like to thank Howard Beck for his guide, challenging questions and support.
I will also give a special thanks to Sem Lampotang for his strong support and decision to
make use of this work for his work without hesitation. I am grateful to Herman Lam for
his encouragement and very helpful comments throughout this work. A special thanks
goes to Abdelsalam Helal for his guidance. I would like thank Markus Schneider for his
support and serving as an advisory committee.
Thanks to all my colleagues at the Database Systems R&D Center who gave me lot
of help. Especially I would like to thank Seema Degwekar and Xu Zhang for helping me
with implementations and Brian Weinrich for proofreading this dissertation. A special
thanks to Sharon Grant for reviewing many research papers even with very short notices.
I cannot thank enough to my parents for giving me life in the first place and their
endless support. I also would like to thank my parents-in-law for their support. Last, but
not least, I would like to give a very special thanks to my wife, Rang-wook Oh, who
always has been there for me and to my lovely daughter Narae for their patience and
understanding all the hard works I went through.
TABLE OF CONTENTS
page
A C K N O W L E D G M E N T S ................................................................................................. iv
LIST OF TABLES ................................. ............. .............. .... ix
LIST OF FIGURES ............................... ... ...... ... ................. .x
ABSTRACT ........ .............. ............. .. ...... .......... .......... xii
CHAPTER
1 IN TR O D U C TIO N ................. .................................. .... ....... .................
2 R E L A T E D W O R K .............................................................................. ............ ... 7
2.1 Cisco's Reusable Learning Object (RLO) Strategy.................. ..................
2.2. ADL's Sharable Content Object Reference Model (SCORM)..............................8
2.3 Flex-EL ............... ........ ......................................................................11
2.4 K now ledgeTree ........................................... ............ ...... .......... 12
2 .5 L 3 .................................................................................................................... 1 2
2.6 NETg ............... .......................................... 13
2.7 M ER L O T ............................................................................................. ............ 13
2 .8 C O R D R A ........................................................................................................ 1 3
3 LEARNING OBJECT MODELS ................................. ............................15
3.1 W hat Is a Learning Object? ................................ ...................... ..... 15
3.1.1 Atomic Learning Object ................................ .................... ............ 15
3.1.2 Composite Learning Object ........................ ........................... .......... 16
3.1.2.1 Activity tree ................................ ........................ ... ... 16
3.1.2.2 Sequencing control modes ........................................................18
3.1.2.3 Learning objective ......................................................... ........ 19
3.1.2.4 A activity action ........................................................... .. ...... 20
3.1.2.5 Activity data ................................ ....................... .... ..20
3.1.3 M eta-D ata .............................................................. 20
3.2 Learning Resources Taxonom y .................................................... ....... 22
3.3 Dynamic E-Learning Service Requirements and Techniques...........................23
3.3.1 Rule-Based Execution M odel ............................................ ....... 24
3.3.2 D ynam ic B inding........................................ ............... ............... 29
3.3.3 M odel-based A ssessm ent ........ ................................................................ 31
3.3.3.1 Assessment specification............. ...... ...............32
3.3.3.2 Model-based objective measure assignment.................................33
4 LEARNING CONTENT MODELING LANGUAGE AND LEARNING
PROCESS MODELING LANGUAGE ........................................ ............... 36
4.1 W hy XML? ................................. .............................. ........ 36
4.2 M eta-m odel ..............................................38...........................
4.3 Learning Content Definition Language ............... ......................................39
4.4 Learning Process Definition Language ..................................... .................42
4.4.1 Non-leaf Activity ........................................... ..... ... ..................43
4.4.2 Leaf A activity ................................. ............ .... ........ .. .............. 48
4 .4 .3 C o n n e cto r ............................................................................................. 5 0
4 .4 .4 E d g e ...............................5 1.............................
5 E-LEARNING SERVICE INFRASTRUCTURE ............................................... 53
5.1 E-Learning Infrastructure for Virtual E-Learning Communities........................53
5.1.1 Roles in a Virtual E-leaming Community...............................................53
5.1.2 Web-Services-Based E-Leaming Service Infrastructure..........................56
5.2 Implementation ............ ....... ....................... ...... ................ 57
5.2.1 Constraints-enabled Web-service Model ....................................... 58
5.2.2 Authoring Tools & LO Repository.................................. ............... 62
5.2.2.1 AL O authoring tool ........................................ ....... ............... 62
5.2.2.2 CLO authoring tool ........................................ ....... ............... 64
5.2.2.3 LO repository ............. ................................ .......... .. .......... 64
5.2.3 Learning Process M management System ................... ................................. 65
5.2.3.1 Learning process execution engine .............. ................................. 65
5.2.3.2 Event-Trigger-Rule server................. ....... ........ ........... 71
5.2.3.3 A ssessm ent com ponent ....................................... ............... 74
5.2.3.4 Constraints-based LO broker ................................. ............... 76
5.2.3.5 U ser interface com ponent...................................... ............... 78
5.2.3.6 W eb user interface................................................. .. .............. .. 78
6 APPLYING LEARNING OBJECT AND DYNAMIC E-LEARNING
TECHNOLOGIES TO SIMULATION-BASED MEDICAL INSTRUCTION ........81
6.1 Instruction on Anesthesia M achines................................ ....................... 81
6.2 Introduction to the Virtual Anesthesia Machine............... ......... ............... 82
6.3 Application to Medical Instruction using VAM............... ......... ............... 84
6.3.1 Encapsulating VAM as Learning Objects.................... .............. 85
6.3.2 Dynamic Execution of VAM Instruction ........................................... 88
6.3.3 Importance of Leaming-Object-Technology-Based Medical Instruction..91
6.4 Chapter Summ ary .................................. .. .. .. ........ ...............92
7 SU M M ARY ......................... .. ......... .......... ..................................94
7.1 Sum m ary and C onclusion.......................................................... ............... 94
7 .2 F u tu re W o rk .................................................................................................... 9 5
APPENDIX
A XML SCHEMA DEFINITION OF LEARNING CONTENT DEFINITION
LANGUAGE .............. ................. ............ ................ .......... .. 97
B XML SCHEMA DEFINITION OF LEARNING PROCESS DEFINITION
LANGUAGE .............. ................. ............ .............. ............ 102
C SAMPLE ATOMIC LEARNING OBJECT AND COMPOSITE LEARNING
O B JE C T ...................................... ................................................. 1 12
C.1 Sample Atomic Learning Object Low Pressure System.............................. 112
C.2 Sample Composite Learning Object Virtual Anesthesia Machine ...............15
D LEARNING OBJECT REPOSITORY AND BROKER WEB SERVICE
DEFINITION LANGUAGE DOCUMENTS....................................................... 129
D. 1 Learning Object Repository Web Service Description.................................... 129
D .2 Broker W eb Service D definition ........................................ ...... ............... 131
E XML SCHEMA DEFINITION OF CLO RUNTIME STATUS MODEL AND
LEARNER INFORM ATION M ODEL................................................................. 135
E.1 CLO Runtime Status Model Definition ................................... ............... 135
E.2 Learner Information M odel Definition ................................... ............... 138
LIST OF REFEREN CES ........................................ ........................ ............... 146
B IO G R A PH ICA L SK ETCH ......... ................. ...................................... .....................151
LIST OF TABLES
Table page
4-1 Execution Stages and Description of Activity Actions in Non-leaf Activity.............45
4-2 Execution Stages and Description of Activity Actions in Leaf Activity...................49
LIST OF FIGURES
Figure page
2-1. C isco 's R L O and R IO .......... ............................................ ................ ........ ....... .. ..
2-2. A Representation of an Activity Tree Composed of Clusters in SCORM .................10
2-3. Conditions and Action/Behavior in SCORM's Sequencing Definition Model.......... 10
3-1. Modeling Constructs in a Composite Learning Object.................. ................17
3-2. Sequencing Control M odes. .............................................. ............................. 18
3-3. Meta-data Attributes in Extended WSDL Document..............................................22
3-4. Taxonomy of Learning Resources and their Relationships........................................23
3-5. Tasks Performed in an Activity Node at Runtime.................. ................................25
3-6. Assignment of Measures to Multiple Objectives for an Assessment at Al ..............34
3-7. Activity Tree with M multiple Assessments ...................................... ............... 35
4-1. Meta-model of Learning Process Definition Language and Learning Content
Definition Language and their Relationships.......................................................38
4-2. D detailed M eta-m odel for L CD L ........................................................................ .. ...42
4-3. Entities in the Meta-model of a Non-leaf Activity in Learning Process Definition
L an gu ag e ............................................................................. 4 7
4-4. Meta-model of Leaf Activity and Related Entities in the Learning Process
D definition Language .................................................. .. ...... .. ........ .... 50
4-5 Meta-model of Connector and Edge in the Learning Process Definition Language...51
5-1. Roles and Relationships among Members of a Virtual E-Learning Community.......55
5-2. Web-service-based E-Leaming Service Infrastructure ...........................................57
5-3. Extended W SDL Information M odel .............................................. ............... 59
5-4. Extended W SDL Document Example ............................ .................................... 61
5-5. Graphical User Interface Frame of the ALO Authoring Tool...................................62
5-6. Graphical User Interface Frame of the CLO Authoring Tool ................. ........... 63
5-7. Architecture of Learning Process Management System.....................................65
5-8. State Transition Diagrams for Learning Process and Activity .............................66
5-9. Relationships between Learning Object Definition and Runtime Objects of a CLO
Instance ............... .. .. ....... ................. ................................67
5-10. Architecture of the Learning Process Execution Engine................................. 68
5-11. Object Diagram of CLO Deployment Process .............. ......................70
5-11. Event and Rule Mechanisms for Coordination and Collaboration.........................73
5-12. Runtime Structure of Assessment Component.................................... ..................75
5-13. Registration and Discovery Process of LO..................................................... 76
5-14. W eb U ser Interface ............... ................. .......................... .... ..... 79
6-1. Construction of LOs using VAM Learning Resources .............. ................86
6-2. W eb User Interface of VAM Instruction...................... ................... ............... 87
6-3. Condition/Action Rules and Dynamic Binding Defined in 'VAM' Instruction
C L O .................................................................................90
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
A WEB-SERVICE-BASED E-LEARNING SERVICE INFRASTRUCTURE FOR
ACHIEVING DYNAMIC AND COLLABORATIVE E-LEARNING
By
Gilliean Lee
August 2005
Chair: Stanley Y. W. Su
Major Department: Computer and Information Science and Engineering
Multimedia data and application systems that are accessible over the Web are
valuable assets for developing instructional materials for teaching, training, problem
solving and decision support. These assets can be used for constructing learning objects,
each of which is a reusable granule of instruction designed to meet a specific instructional
objective. In order to search, access, and deliver the learning objects, an infrastructure for
the registration, discovery, and invocation of these objects is required. Also, there is a
need to develop infrastructural technologies to enable general Internet users, not just
students and educators, to engage in life-long instruction and learning. These
technologies should allow the establishment of many virtual e-learning communities,
each of which is formed by people with a common interest in exchanging knowledge
useful for problem-solving and decision support in a specific problem domain.
In this work, we model distributed and sharable learning resources by two types of
Learning Objects (LOs): Atomic Learning Object and Composite Learning Object. Both
types of LOs are uniformly published as Web-services in a constraints-based Web service
broker. XML-based languages for modeling these two types of LOs are defined. The
languages also serve as the interchange formats for transferring and sharing LOs.
We have developed an e-learning service infrastructure that facilitates the authoring,
registration, discovery, storing, and invocation of LOs. The infrastructure is built on top
of an extended Web-service framework, which leverages the constraints-based Web
service broker and distributed LO repositories to make LOs searchable, sharable and
reusable. A set of distributed software components has been developed to provide e-
learning services. It includes authoring tools for constructing LOs, and a Learning
Process Management System for the registration, management and delivery of LOs and
the assessment of learners' comprehension. The techniques used to achieve active,
adaptive, flexible and customizable execution of learning processes such as dynamic
binding to LOs, rule-based execution of learning processes that makes use of the profiles
and runtime status models of learners, and model-based assessment are presented. The
roles and functions of virtual e-learning community members are also discussed.
The e-learning technologies we developed have been applied to encapsulate the
instructional materials of the Web-based Virtual Anesthesia Machine and the simulator as
LOs, deliver instructional materials to learners, use the simulation system for practice,
and perform assessment. Several techniques, such as retrying or disabling activities
depending on a learner's profile, performance and needs, dynamic binding to suitable
learning objects, optional assessment presentation, and the support of different learning
approaches by customization of sequencing control modes, have been applied to achieve
personalized learning.
CHAPTER 1
INTRODUCTION
The advent of Internet and Web technologies has enabled organizations and
individuals all over the world to place enormous quantity of multimedia data (texts, Web
pages, graphs, images, Web simulation, video and audio) on the Web. Increasingly, Web-
based application systems are also made accessible to Web users through browsers.
These data and application systems are valuable resources for constructing instructional
materials for teaching, training, problem solving and decision support. One approach to
make use of these distributed, heterogeneous data and application systems is to apply the
object-oriented technology and "wrap" them as distributed objects. These distributed
objects can then be used as learning assets to compose learning objects (LOs) for
instruction and training purposes.
In recent years, there have been a number of initiatives in developing and
standardizing Web-based learning technologies. The Advanced Distributed Learning
Initiative [ADL, 2003], the IMS Global Learning Consortium [IMS, 2004], and the Open
Knowledge Initiative [Eduworks & OKI Leadership, 2002], are a few examples.
The Sharable Content Object Reference Model (SCORM) [ADL, 2004a] is a
reference model initiated by the Advanced Distributed Learning (ADL) program of the
Department of Defense (DoD) and the White House Office of Science and Technology
Policy (OSTP). According to the SCORM's specification, it is envisaged that Internet
users and heterogeneous LMSs would use the Web as a universal platform for accessing
and launching sharable content objects and for establishing close communication,
interaction and coordination among content object developers, course authors, users, and
administrators. To realize this vision, sharable content objects must be durable,
interoperable, accessible and reusable. In order to meet these requirements, it is necessary
to have a uniform way of modeling, not only learning resources, but also heterogeneous
learning tools and LMSs, as well as an information infrastructure to enable the
interoperation and sharing of their contents and functionalities. Also, the aggregation
model that defines the learning sequence or process has to be flexible, adaptable and
customizable to meet different learners' needs and learning contexts. Our work to meet
the above needs is consistent with the vision and goals of the ADL Program.
There are systems such as MERLOT [MERLOT, 2004], MIT OpenCourseWare
[OCW, 2003], CLOE [CLOE, 2004] and EOE [The EOE Foundation, 2004], which offer
repositories of learning assets to virtual learning communities. Each learning asset is
described by metadata and a Universal Resource Locator (URL), which is searchable
either by categories or by keywords using a search engine. These systems function more
like directories provided by digital libraries. Users of these systems have to manually
search the directories and access the learning assets that are "pointed to" by the URLs
stored in the directories. The above systems do not provide externally accessible
programming interfaces for accessing the contents of the directories and the associated
learning assets. Nor do they provide facilities for assembling instructional materials out
of the learning assets available in the Web, delivering content to users based on their
needs and profiles, and assessing users' understanding of the delivered content.
On the other hand, there are commercial instructional management systems (IMSs)
such as WebCT [WebCT, 2005], Blackboard [Blackboard, 2005], and eCollege [eCollege,
2005], and research systems such as LON CAPA [Laboratory for Instructional
Technology in Education, 2004] and Sakai [Sakai Project, 2005], which are designed to
support course development, deliver contents to students, perform assessment, and
provide tools for course management. These systems are designed and implemented for
creating, teaching and managing courses and their targeted users are instructors, students
and course administrators of educational institutions.
The existing systems that provide directory accesses to learning assets and the
existing instructional management systems represent two ends of a spectrum. Useful
systems can be built in between these two ends. For example, a system may provide tools
for Internet users, who are not necessarily course developers or instructors, to design,
develop and share LOs. LOs do not have to be instructional materials for formal courses.
They can be small granules of information content that are useful to some learners and
contain practice and assessment items to exercise and assess the learners' comprehension
of the content. The learners do not have to be students who take courses in schools or
receive some formal training in business and government organizations. They can be
ordinary Internet users who need help in various decision-making and problem-solving
situations. Since such a system is not a full-fledged IMS, it will be simpler and easier to
develop and maintain than IMSs.
Our vision is to build an e-learning service infrastructure, which provides tools for
authoring learning objects, facilities for aggregating and sequencing of small granules of
learning objects to form reusable composite learning objects, and a Learning Process
Management System for learning object registration, management, delivery and
assessment. The e-leaming service infrastructure will enable the global participation of
Internet users in life-long instruction and learning. Internet users with similar interest in a
specific subject or domain will be able to form a virtual e-learning community to share
their knowledge. This is consistent with the vision of "building the technological
infrastructure to support dynamic, ad-hoc communities of lifelong learners who interact
within an environment of learning objects through a creative blend of advanced
computing technologies, high performance networks, authoring and collaboration tools"
stated in York et al. [2002].
Internet users have much more diverse backgrounds than students in schools. A
learning process management system that delivers identical contents to all learners cannot
accommodate diverse Web users' needs. Web-based learning should accommodate
learners' different profiles, which include users' backgrounds, competencies and interests.
A Learning Process Management System must be flexible, adaptive and customizable in
order to satisfy heterogeneous learners' needs.
Learning is largely a social activity; and even the most well developed multimedia
interactive materials lack the flexibility of human interaction [Lewis, 2002]. Therefore, a
Learning Process Management System that facilitates interactions and collaborations
among the people involved in learning will enhance the learning experience.
This work focuses on the creation of an Web-services-based learning, problem-
solving and decision-support environment, in which Internet users (not limited to
educators) who are knowledgeable in any subject domain (e.g., how to purify water, how
to care for a patient with a certain disease, etc.) can participate in a global virtual
community to create, compose, and evaluate distributed, reusable and assessable
multimedia LOs for instructional purposes; and Internet users who engage in lifelong
learning, problem-solving and decision-making can request, access and study the contents
of these LOs, be assessed of the knowledge gained, and engage in collaborative learning
with other learners and content producers.
The e-learning virtual communities envisioned in this work are different from the
existing virtual communities on the Internet. Learning in the virtual e-leaming
communities is based on reusable, assessable, and programmatically accessible e-leaming
objects. LOs are stored in distributed repositories of content providers to avoid the
bottleneck problem of a centralized repository and to give content providers more
authority and freedom to manage their LOs. These objects are designed and evaluated by
content producers and content evaluators with the support of user interface tools. They
are delivered and managed by the Learning Process Management System, which consists
of a Learning Process Execution Engine (LPEE) for processing learning objects, an
Event-Trigger-Rule (ETR) Server [Lee, Su & Lam, 2001] and other supporting
components for achieving collaborative, adaptive, customizable and flexible e-learning.
To summarize, our vision is the creation of a sustainable e-learning infrastructure
and the development of advanced e-learning technologies to enable the global
participation of Internet users in life-long instruction and learning. Internet users with
similar interest in a specific subject or domain will be able to form and operate their
virtual e-learning community to share their knowledge captured in reusable, compose-
able and assessable learning objects and use them to aid decision-making and problem
solving. The contribution of this work is in the development of an extended Web-services
infrastructure and dynamic and collaborative e-leaming technologies essential for the
realization of our vision. The developed infrastructure and technologies are in forms of
learning object models, learning object authoring tools, learning object definition
languages, user interface tools, information and runtime status models of learners,
distributed e-learning service architecture, event-trigger-rule-based execution of learning
processes, and model-based assessment.
This dissertation is organized as follows: Related works are discussed first. Then,
two learning object models for defining what we call Atomic Learning Object (ALO) and
Composite Learning Object (CLO) are presented in Chapter 3. The languages used to
define ALOs and CLOs are then described in Chapter 4. Next, Chapter 5 describes an e-
learning service infrastructure for virtual e-learning communities. The implementation of
the infrastructure including tools and software system components are also discussed.
Chapter 6 discusses an application of our e-learning technologies to encapsulate the
instructional materials of the Web-based Virtual Anesthesia Machine and the simulator as
LOs, deliver instructional materials to learners, use the simulation system for practice,
and perform assessment. A summary and the follow-up R&D efforts are given in the last
section.
CHAPTER 2
RELATED WORK
Effective delivery of learning contents to learners of varied backgrounds is one of
the goals of e-learning. Organizing learning contents is an importance task for achieving
this goal. There are several research and standardization efforts in developing learning
process models. SCORM's Content Aggregation Model [ADL, 2004b] and Cisco's
Reusable Learning Object [Termaat et. al., 2003] are two popular models that organize
learning objects for effective delivery. The structure of learning content specifies the
delivery sequence of the content. A tree structure, which is commonly used to present the
'table of contents' of a book or training manual, is generally used to model a learning
process. A tree node represents a granule of learning content to be presented to learners.
It can represent a course, module or lesson. The tree structure is used in SCORM, Cisco's
reusable learning object (RLO), CLI Virtuoso [Cisco Learning Institute, 2004],
KnowledgeTree [Brusilovsky, 2002], and KOD [Sampson, 2002].
2.1 Cisco's Reusable Learning Object (RLO) Strategy
Cisco's Reusable Learning Object (RLO) Strategy is a scheme for configuring LOs
that can be used in a learning process or learning management. It recognizes two levels of
LOs: Reusable Information Objects (RIOs) at the bottom level and RLOs at the top level.
Each RIO contains content items, practice items and assessment items related to a
learning objective. These items are stored in a RIO so that the consistency among content
delivery, practice and assessment can be maintained when the RIO is used in different
contexts. A larger structure called RLO can be formed out of RIOs. It can have an
overview, a summary, an assessment, and five to nine (7+2) RIOs. The assessment in an
RLO is simply a collection of the assessment items taken from the RIOs that form the
RLO. When an RLO is delivered, the content items in the RIOs are presented, and then
the assessment items randomly chosen from its RIOs are given. Figure 2-1 shows a RIO,
an RLO and their relationship.
Although an RLO can be meta-tagged, it lacks a meta-data specification that
describes RLOs and RIOs. A concrete specification of meta-data is required especially
for automated discovery and dynamic binding purposes. In our work, we use an Object-
Oriented Constraint Specification Language (OOCSL) introduced in Su et. al. [2001] to
specify the meta-data of a learning object.
A B
Figure 2-1. Cisco's RLO and RIO. A) Reusable Information Object. B) Reusable
Learning Object
2.2. ADL's Sharable Content Object Reference Model (SCORM)
SCORM is a reference model, which is intended to provide a standard framework
for building learning systems to enable the reuse and sharing of learning contents
developed by different authoring tools. Since our learning process model makes use of
some ideas presented in SCORM's sequencing and navigation model, we describe
SCORM's model in more details below.
In SCORM, there are two types of specialized instances of resources: Asset, and
. \,/lablle Content Object (SCO). Asset is a learning content in its most basic form. Assets
are electronic representations of media, texts, images, sound tracks, web pages, chat
sessions, assessment objects or other pieces of data that can be rendered by a Web client
and presented to a learner [ADL, 2004b]. SCO represents a collection of one or more
Assets packaged as a single launchable resource that can communicate with a Learning
Management System (LMS) using its runtime data environment.
SCORM's Sequencing Definition Model [ADL, 2004a] is a rule-based sequencing
model, which supports adaptive execution of an activity tree using sequencing rules,
rollup rules, and the status of learning objectives. An activity tree is a hierarchical
structure of learning activities and their corresponding learning contents. A cluster,
composed of a parent and its child activities, is the basic building block for sequencing.
An activity tree is composed of a set of clusters. The parent activity specifies information
about the sequencing strategy for navigating through its child activities. Figure 2-2 shows
the structure of an activity tree and its clusters in SCORM's Sequencing Definition
Model. An activity can have an associated resource such as SCO resource or Assets
resource, or can be made up of additional sub-activities consisting of other clusters.
In SCORM, learning objectives represent a set of locally and globally scoped data
items, each with a satisfaction status and a satisfaction measure. Activities may have
more than one associated local objective and may reference multiple globally shared
objectives. The Tracking Status Model records the status of progress associated with
objectives, activities, and attempts of activities.
10
Cluster
Course
t----------- /,-- T
II I I
Module 1 Module 2 Module 3
I------------------- ---L--------
-Lesson 01 Lesson 01 Lesson 01
II I I
Leso 0 Lesson 02 LeSSO 2 Lesson 02
-- II I I Cluster -
CluCluster I ste
Cluster -
Lesson 03 Chapter 1
-- Lesson 04 Chapter 2
Cluster Cluster
Figure 2-2. A Representation of an Activity Tree Composed of Clusters in SCORM
SCORM's Sequencing Definition Model makes use of a rule-based sequencing
model, which adopted the IMS Simple Sequencing Specification [IMS, 2003]. A set of
rules is defined for an activity, and is evaluated at specific times during the delivery of an
activity. A sequencing rule is in the form of if[condition] then [action/behavior]. It
consists of a set of conditions and an action/behavior. The action is executed when the
conditions of the rules are evaluated to True. The conditions of a sequencing rule are
based on the Tracking Status Data. Figure 2-3 shows the possible conditions and
actions/behaviors in sequencing rules.
Precondition Postcondition
Actions Actions
Conditions Skip Exit Parent
Activity Progress Disabled Exit All
Satisfied Known
Known
Objective Status Attempted Hidden From Choice Retry
Known
Known A Stop Forward
IF Objective Measure Attempt Limit Then Stp raverwa Retry All
(not) Known Exceeded
Objective Measure Time Limit Exceeded Ignore (default) Continue
Greater Than
Objective Measure Outside Available Previous
Less Than Time Range Exit Actions
Ignore (default)
Completed Always (default) Exitgnore (defa
Ignore (default)
Figure 2-3. Conditions and Action/Behavior in SCORM's Sequencing Definition Model
SCORM's Sequencing Definition Model provides the desirable set of features
described above; however, it has a few limitations. First, the definition of learning objects
in SCORM does not explicitly distinguish content, practice, and assessment items as in
Cisco Systems' RLO, which is adopted in our work. Additionally, we make use of meta-
data and their constraints specified in each LO to achieve more accurate, dynamic
binding of requests to LOs. Second, non-leaf activities in an Activity Tree neither present
contents nor perform assessment in SCORM. We believe that it is useful to allow a non-
leaf node to present an introduction and/or a summary of the contents covered by its child
activities. Assessment items for testing the integrated knowledge of the contents
presented by the child activities should also be allowed. Third, there is no model or
specification for conducting assessments in the SCORM sequencing definition model,
hence the coverage of an assessment is decided entirely by a SCO bound to the activity
instead of the information specified in its sequencing model. As a result, a change of
assessment coverage in an activity tree requires a structural change of the learning
process and the authoring or modification of an SCO to conduct the new assessment. This
may require a substantial amount of time and effort. In this work, we introduce a model-
based assessment technique to facilitate flexible assessments.
2.3 Flex-EL
A workflow process model represented by a directed acyclic graph (DAG) is
employed in Flex-EL [Lin, 2002]. The model can specify the interaction, coordination
and collaboration among the individuals involved in learning and also the sequencing of
contents to be presented to learners. The system uses conditional transitions expressed in
a workflow model to achieve the flexibility in selecting learning pathways that match
each learner's individual pace and learning style. Different from Flex-EL, we model a
learning process by an activity tree, and use sequencing rules and rollup rules as in
SCORM. The action part of the rules can change the sequencing paths by allowing
learners to retry, hide, and skip activities when the conditions of the rules are met. In
addition, we allow the structure of a learning process or sequencing control modes to be
customized at runtime by invoking some rules to check a learner's profile and progress.
2.4 KnowledgeTree
The adaptive Web-based educational system, KnowledgeTree, introduced by
Brusilovsky [2002], separates learning materials into primary materials for average
learners and additional materials for learners with different learning styles and knowledge.
The system uses learning goals, preferences and knowledge of individual learners to
select the most appropriate learning materials. Our dynamic binding of requests to LOs is
similar to their idea of binding learning materials at runtime. However, we leverage on
the standardized Web-services technology for the registration and discovery of LOs and
add meta-data and constraints in LO specifications.
2.5 L3
L3 [Altenhofen, 2002] separates learning processes from strategies, that is,
navigation rules, and introduces meta-tags for describing the knowledge type of learning
activities and the inter-relationships among the activities. A strategy chosen at runtime
and the meta-tags of learning activities enable adaptive navigation of a learning process.
Different from their work, our learning process model allows sequencing rules and rollup
rules to be attached to each activity at both design time and runtime in order to achieve
dynamic properties in a learning process execution.
2.6 NETg
NETg [Thomson, 2005] is a commercial learning management system. With pre-
assessment technique, NETg can selectively deliver contents that a learner needs to take
according to the assessment result. However, different from this work, they do not make
use of post-assessment results and learner profile for adaptive delivery. In NETg, the
assessments can be taken anytime a learner wants, and the deliverable learning contents
are written in a proprietary format instead of the general Web document format.
2.7 MERLOT
MERLOT (Multimedia Educational Resource for Learning and Online Teaching)
[MERLOT, 2004] is a free and open resource repository designed primarily for faculty
and students of higher education. MERLOT is also a community of people who are
involved in education. Community members may contribute materials, assignments and
comments. Meta-data-based search and category-based search methods through a Web
browser are provided, and both require human intervention. Learning resources can only
be accessed through the URLs provided by the system. It does not provide an access
mechanism for an external software component to access the resources.
2.8 CORDRA
CORDRA (Content Object Repository Discovery and Registration/Resolution
Architecture) [LSAL, 2004] is a reference model for content repositories, brokering of
contents, and APIs for content registration, discovery and a standard mechanism to access
repositories. Its key capabilities include the following:
* Content can be made widely available.
* Content can persist outside of the realm of a single course.
* Content can be discovered.
* Once discovered, there will be standard mechanisms to access the content
14
* Content can be managed and deploying organizations will be able to tailor
management to their needs.
* Implementation of the model will be open and flexible.
Our design requirements and functionality of the implementation of content
repository and constraints-based broker are consistent with CORDRA's functional
requirements. We leverage standard Web-service technology and Web-service registry
and make some extensions to them for the registration, discovery, and access of learning
objects.
CHAPTER 3
LEARNING OBJECT MODELS
On the Web, there is an enormous amount of multimedia data and application
systems that can potentially be used for constructing instructional materials for teaching,
training, problem-solving and decision support. The object concept in Object-Oriented
Programming can be applied to encapsulate these materials as objects for their use and
reuse.
3.1 What Is a Learning Object?
A Learning Object (LO) is a granule of instruction designed to meet a specific
instructional objective. We model an LO in terms of content items, practice items,
assessment items needed to cover a subject of learning, provide practice items and
conduct an assessment. Additionally, we specify its meta-data as introduced in Termaat et
al. [2003] and Heins & Himes [2002]. By including these three types of learning items in
an LO specification, the author of the LO can ensure that practice and assessment items
presented to learners are consistent with the presented content items. After an LO is
authored, we publish the LO as a Web-Service in a Constraints-based Web service
Broker for interoperability and ease of access.
3.1.1 Atomic Learning Object
An Atomic Learning Object (ALO) is a small granule of instruction, which
contains some content items for teaching learners a problem-solving skill or providing
them the knowledge needed in a decision-making situation. An ALO can optionally
include practice and assessment items for exercising and assessing learners' acquired
skill or knowledge. In addition to these learning items, meta-data that describes
properties and constraints of the LO, and a delivery sequence of content and practice
items compose an ALO.
3.1.2 Composite Learning Object
ALOs by themselves will have limited use unless they can be used to compose
larger granules of instruction. A Composite Learning Object (CLO) is an aggregation
model defined as a tree structure of activities with sequencing information and
sequencing rules. The following subsections discuss the specification of an activity tree
including sequencing control modes, learning objectives and activity data.
3.1.2.1 Activity tree
When a reader selects a book, he/she usually takes a look at the table of contents of
the book to view the contents and the overall structure of the book. An activity tree in a
CLO follows the tree structure of the table of contents to aggregate deliverable contents
and defines the delivery sequence of the contents. An activity tree is designed with its
modeling constructs such as activities, connectors, and edges.
S Activity: A unit of processing in the delivery of a CLO instance. An activity can
represent a granule of instruction such as a lesson, module, chapter, part, and so on.
Depending on its location in an activity tree, an activity is either a leaf or a non-leaf
activity. A non-leaf activity can include content, practice and assessment items in its
specification. A content item in a non-leaf activity can provide an abstract and/or an
introduction to learners before the sub-tree rooted at the non-leaf activity is
delivered. A summary of the contents of a sub-tree can also be presented by a
content item. Assessment and practice items in a non-leaf activity can be designed
to assess learners' abilities to integrate the concepts and contents presented in its
child activities. A leaf activity can bind to a specific ALO or CLO statically, or it
can contain constraints for dynamic binding to an ALO or CLO that is suitable to a
learner. Additionally, a set of rules can be included in a CLO specification and be
used to alter the navigation path and the structure of an activity tree based on each
individual learner's profile (i.e., background, age, preference and competency) and
progress. In an activity, learning objectives and activity data can be declared and
specified. The sequencing rules and the rule-based execution model will be
discussed later in this chapter.
* Connector: A modeling construct that bridges a parent activity and its child
activities. It includes Sequencing Control Modes of the child activities, which will
be discussed in the next section.
* Edge: A directed edge that connects a parent activity to a connector, and a
connector to a child activity.
Figure 3-1 shows an example CLO, 'Relational Model', composed of the modeling
constructs discussed above.
-el
Relational Model
I--I
E 4,------ Edge
I I
I--I
Non-leaf Activity
E E E E
-___---___
I II I S I
History Terminology Referential Integrity Relational Langsi
Leaf Activity C" --- Connector
E E E
E _
Relational Algebra Relational Calculus Other Languages
Figure 3-1. Modeling Constructs in a Composite Learning Object
18
3.1.2.2 Sequencing control modes
Sequencing control modes (SCMs) adopted from SCORM's Sequencing Definition
Model defines the sequencing behavior among child activities. SCMs include Choice,
Forward-Only and Flow modes. Each mode can be set to True or False, and multiple
modes can be enabled to specify combinations of these modes. Choice SCM enables the
user to choose child activities in any order. Forward-Only SCM permits traversal through
its child activities only in the forward direction. Flow SCM allows sequencing through
the child activities in forward and backward directions. The sequence of child activities in
Flow and Forward Only SCM is defined in a connector. Figure 3-2 illustrates the three
sequencing control modes.
Parent Activity
Choice
Activity 1 Activity 2 Activity 3 Activity 4
A
Parent Activity
Flow
Activity 1 Activity 2 < Activity 3 < Activity 4
B
Parent Activity
,
----- ----- -- C
Figure 3-2. Sequencing Control Modes. A) Choice. B) Flow. C) Forward Only.
Unlike SCORM, which defines SCMs in a parent activity of a cluster, we specify
them in a connector. By separating the control information (i.e. SCMs) from the activity
specification, runtime customization of a learning process instance can be achieved more
easily.
3.1.2.3 Learning objective
A learning objective, adopted from SCORM's Sequencing and Navigation model
[ADL, 2004a], is a data structure whose data conditions can either be marked as
'Satisfied' or 'Unsatisfied' to indicate if a learning objective has been or has not been met.
Learning objectives are declared in an activity. They can be prefixed by the ID of
the activity (i.e., 'activitylD.objectivelD'), thus can be unambiguously referenced from
any activity of a CLO. To reference a locally declared objective, activityld does not be
prefixed. A learning objective includes the data fields such as 'Objective ID (String)',
'Satisfied by Measure (Boolean)', 'Minimum Satisfied Measure (decimal value between
0.0 and 1.0)', and 'Contributes to Rollup (Boolean)'. An objective whose 'Contributes to
Rollup' is True is referred to as primary objective. Each activity is required to have a
primary objective, and the satisfaction status of the primary objective represents the
satisfaction status of the activity. At runtime, a learning objective will have a Boolean
value for its satisfaction status and a numerical value for its normalized score, that is.
measure.
There are two ways to satisfy or unsatisfy an objective: First, when an objective's
satisfied by measure field is True, a measure (score of an assessment) will be recorded as
the result of an assessment. If the measure is equal or above its minimum satisfied
measure, then the objective will be set to 'Satisfied'. Otherwise, it will be set to
'Unsatisfied'. Second, when an objective's satisfied by measure field is False, the
objective has to be set to 'Satisfied' or 'Unsatisfied' by a condition-action rule. For
example, in order to satisfy an activity when a learner has taken all the content and
practice items, an after-post-assessment rule "if [true], then [Satisfy this activity]" can be
specified. Details of sequencing rules and the stages of activity processing will be
discussed later in this chapter.
3.1.2.4 Activity action
Activity action is a set of simple Java instructions executed at specific execution
stages during activity processing. These instructions may check a learner profile and the
runtime data model, and assign meaningful values to activity data. Details about their use
will be discussed in Chapter 4.
3.1.2.5 Activity data
Activity data are variables declared in an activity. Their values represent the status
of the activity, which are used by rules. The data type of each datum can be any one of
the simple data types of the Java language such as integer, float, boolean, string, and so
forth.
3.1.3 Meta-data
Meta-data of an LO is the data that describes the LO. The meta-data is expressed in
terms of a set of attributes and the constraints associated with these attributes. The
attributes include Title, Subject, Author, Keywords, Target age, Media format, Language,
Cost, Quality Rating, and so on, as in MERLOT, CLOE, and IEEE LOM standard [IEEE
LTSC, 2005]. These attributes can be categorized into two types. Static attributes such
as Title, Subject, Keywords, Language, Target age and Media type are attributes whose
constraints are specified by authors. Dynamic attributes such as Usage number, Average
score and Quality rating are attributes whose constraints are recorded and updated by
evaluators (a role played by qualified individuals in a virtual community) and the
Learning Process Management System. Constraints can be categorized into attribute
constraints (i.e., the legitimate values of these attributes), and inter-attribute constraints
(i.e., their value relationships). An attribute constraint has the following form:
* attribute-Id keyword value range enumeration
where attribute-Id is the identifier of an attribute, keyword can be either a
comparison operator '=', '!=' <', '<=', '>' or '>=', 'RANGE' or
'ENUMERATION'. In case the keyword is a comparison operator, value would
specify the data value of the attribute in a simple predicate expression. If the
keyword is 'RANGE', a range specification denoted by [value..value] would
specify the value range that the attribute can have. In case of 'ENUMERATION',
an enumeration specification denoted by [value, value,.., value} would specify a list
of possible values of the attribute.
An inter-attribute constraint is defined as follows:
* If attribute-constraints then attribute-constraints
where attribute-constraints is a Boolean expression in which each term is an
attribute constraint.
For example, an attribute constraint of an LO specified by 'Target age = [3..8]'
specifies that the LO is to be used by children between 3 and 8 years old. An LO that is
free of charge to people from a developing country can be specified by an inter-attribute
constraint 'If learner.country e developing country, then cost = 0'. The meta-data
constraints provided by LO producers are used for the registration and discovery of LOs
[Lee, Zhang & Su, 2004].
LOs developed for different application domains may require different meta-data
attributes. In order to accommodate different applications, meta-data attributes must be
customized and/or expanded to meet different application needs. We specify meta-data
attributes in an extended WSDL document, which is used by the LO authoring tools and
LO search tool, and allow the user to specify meta-data constraints. A part of the
extended WSDL document that describes the names and data types of the meta-data
attributes used in our current system are shown in Figure 3-3.
LO Type
string
Title
string
Subject
string
Author
string
Description
string
Keywords
string
Material Type
string
Media Format
string
Primary_audience
string
Language
string
Difficulty
string
Workload
string
Target age
integer
Level of detail
string
Studyperiod
duration
Cost
float
Figure 3-3. Meta-data Attributes in Extended WSDL Document
3.2 Learning Resources Taxonomy
We distinguish three types of learning resources in a taxonomy of resources. The
lowest level of resources is called learning asset, which represents the most basic
resources available on the Web and can be presented to a learner through a Web browser.
A learning asset can be a simple file, such as a text, image, audio/video clip or plug-in
files such as Flash, or a complicated web page that references multiple learning assets.
The next level of resources is the Atomic Learning Object (ALO) in which each of the
learning items includes a learning asset. The highest in the taxonomy is the Composite
Learning Object (CLO), which is an aggregation model defined as a tree structure of
activities with sequencing information. As discussed, a leaf activity can bind to either an
ALO or a CLO for the delivery of learning contents, practice and assessment. Figure 3-4
illustrates the taxonomy and reference/binding relationships among learning resources.
CLO
Non-leaf Activitv /Meta-data,
Intro, summaryI s tra
Practice items
Assessment itemsVidLeaf Lea
Learning Assets-L g ALO R
Image CEEE~^ -Intro, summary
As dImaged, Inteet users have much more diverse backgrounds and profiles than
Practice items
Video clip Assessment items
Audio clip Meta-data &
pageB Constraints
-- Binds to
Figure 3-4. Taxonomy of Learning Resources and their Relationships
3.3 Dynamic E-Learning Service Requirements and Techniques
As discussed, Internet users have much more diverse backgrounds and profiles than
students in schools. Therefore, Web-based learning has to be dynamic to accommodate
learners' different backgrounds, competencies and interests. To meet this requirement,
learning object services must have the following dynamic properties: active,flexible,
adaptive, and customizable. They should be "active" in that operations on them can
automatically trigger rules to enforce policies and constraints, coordinate learning
activities, and notify peers and content producers to achieve collaborative e-leaming.
They should be "flexible" in that small granules of LOs and multimedia assets can be
used to flexibly compose larger objects, and a request for an object can be flexibly bound
to a suitable LO dynamically. They should be "adaptive" in that activities can be retried,
skipped, exited, enabled or disabled, and sequencing control modes can be changed at
runtime to adapt to the learner performancee (i.e., assessment results) and interaction
with the system. They should be customizablee" in that the structures, enabled/disabled
status of activities and sequencing control modes can be customized to suit individual
learners before the processing of a CLO by using the information provided in learners'
profiles. A learner's profile contains a learner's identification, demographics,
accessibility, background and preference information such as date of birth, name, address,
preferred language and learning style, the type of internet connection, disability, interests,
learning history, learning objective, and others. Details of the learner profile model will
be discussed in Chapter 5. The following sections describe the techniques and
mechanisms we have developed to satisfy these requirements.
3.3.1 Rule-Based Execution Model
In this work, we use a rule-based execution model to process a CLO in order to
achieve dynamic, flexible, adaptive and customizable e-learning services. The execution
model can guide learners through different learning paths that lead to effective learning
experience by invoking sequencing rules during the learning process execution. A CLO
instance refers to a runtime instance of a CLO being enacted for the delivery purpose. A
CLO instance is processed by the Learning Process Management System, which contains
a Learning Process Execution Engine (LPEE), an Event-Trigger-Rule (ETR) Server [Lee,
Su & Lam, 2001] and a few other server components.
Drill-down Notified Notify Roll-up
0 Pre-activity Event After-Post-assessment Event+
10 Post-assessment (if specified)
0* Pre-assessment (if specified) t
Condition- Present Practice Items Condition-
Action After-Pre-assessment Event Action
Rules Present Summary Content item Rules
~-- Present Overview No No
S Content item Yes Chld tvt t v
*Drill-down Event R
Notify Drill-down Roll-up Notified
A
Drill-down Notified Notify Roll-up
Pre-activity Event
Bind to a Learning Object
S Pre-assessment (if specified)
Condition-
Action 4. After-Pre-assessment Event
Rules
Present Content/Practice items
4 Post-assessment (if specified)
SAfter-Post-Assessment Event
h B
."""> Event Posting
SExecution Flow
Figure 3-5. Tasks Performed in an Activity Node at Runtime. A) Non-leaf activity. B)
Leaf Activity.
Each enactment of a CLO by the LPEE would process the activities of an activity
tree by following a sequencing order determined by the followings:
* A pre-order traversal of the activity tree (drill-down and roll-up process).
* Sequencing control modes (Choice, Flow and Forward Only).
* Activation of sequencing rules to retry, skip, exit, enable or disable activities.
* Learner's navigation requests at runtime, for example, exit, next activity, previous
activity, choose an activity, continue, and so on.
As shown in Figure 3-5 (A), sequencing rules are executed during the processing of
activities. A sequencing rule is a condition-action rule having the structure, "if
[condition] then [action] else [alt-action]". The major differences between the
processing of leaf and non-leaf activities are that drill-down and roll-up rules do not exist
in a leaf activity, and an LO is bound to a leaf activity after the pre-activity event is
posted, as shown in Figure 3-5 (B).
During the execution of an activity, events are posted to notify the ETR Server that
certain stages of execution have been reached. The ETR Server would invoke sequencing
rules associated with the events. As shown in Figure 3-5, a sequencing rule is invoked by
an event with the same name. For example, a Pre-activity event activates a Pre-activity
sequencing rule. The detailed description of the ETR Server will be given in Chapter 5.
An activity may contain the following sequencing rules:
1. Pre-activity rule Processed before the enactment of an activity. Learner's profile
or the status of another activity can be referenced in the rule for possible adaptation
or customization.
2. After-pre-assessment rule Processed after the pre-assessment task. The result of
the pre-assessment can be evaluated for any possible adaptive sequencing.
3. Drill-down rule Processed before leaving the activity to go down the activity tree
to process its child activities. Sequencing control modes of the connector between
the activity and its children can be set adaptively at this stage by checking the
learner's profile and runtime status. Drill-down rules are applicable only to non-leaf
activities.
4. Roll-up rule Processed when an activity is revisited after each of its child
activities is finished. A roll-up rule is in the form of if[Child_activity_set
(Minimumset) Condition_set] then [action] else [alt-action]. The
Child activity set specifies a qualifier (All, Any, None, At Least Count, or At
Least Percent), which defines the set of child activities referenced during the
evaluation of the condition part of a rollup rule. If "At Least Count" or "At Least
Percent" is used, it is required that Minimum set is specified. Minimum set
represents the minimum count or the percentage of child activities that must be
evaluated. The condition set of the rollup rule contains "Satisfied", "Attempted",
"Completed", "Objective Status Known", "Objective Measure Known", "Activity
Progress Known", "Attempt Limit Exceeded", and "Never" as in SCORM's
Sequencing and Navigation [ADL, 2004a]. For example, a rollup rule "if[At Least
Count (2) Satisfied] then [Satisfy this activity]" states that if any two of the child
activities are satisfied, then the activity (parent) is satisfied. Rollup rules are
specified in a parent activity.
If there are one or more child activities yet to be processed, sequencing control
modes will determine a set of available child activities for the learner's choice.
Roll-up rules are applicable only to non-leaf activities.
5. After-post-assessment rule Processed after a post assessment task to check the
result of the post-assessment. The result can be used for any possible adaptive
sequencing.
The condition part of each sequencing rule may check a learner's profile and the
progress status at each stage of activity processing. Thus, an appropriate action can be
taken to accommodate a specific learner or a category of learners in different learning
contexts. For example, the execution of a sequencing rule can change the sequencing
control mode and/or the control structure of the current instance of an activity tree to
dynamically adapt to the learning context and the learner's profile and progress. The
action part of a rule may specify "Skip Activity", "Exit Activity", "Retry Activity",
"Enable/Disable Activity", "Incomplete Activity", "Satisfy/Unsatisfy Activity",
"Satisfy/Unsatisfy objective", "Hide Activity from Choice", "Pre/Post-Assess",
"Remove/Add Sequencing Control Mode of a Connector", and so forth, so that the
delivery sequence of activities can become adaptive, flexible and customizable.
A pre-activity rule specified in a root activity of a CLO can be activated for
customization of the CLO instance. For example, a pre-activity rule "If[learner 's goals
do not include 'database administrator' or 'DBA'], then [Disable activity
'installing DBMS']" would check the learner profile to see if his/her learning goal is to
become a database administrator. If not, 'installing DBMS' activity that delivers learning
items related to the details of a DBMS installation will not be presented to the learner.
Before an after-pre-assessment rule is invoked, the result of the pre-assessment will be
available. An after-pre-assessment rule "If [pre-assessment objective is 'Satisfied'], then
[Skip this activity would allow the current activity to be skipped when the pre-
assessment result is satisfactory. A drill-down rule of a non-leaf activity specified as "If
[learner prefers 'Receptive learning style], then [Set Sequencing Control Mode to
'Forward Only '], else [Set Sequencing Control Mode to 'Choice 7'" would not allow a
learner who prefers receptive learning style to freely choose the next child activity. Rules
as illustrated above can enable the Learning Process Management System to deliver
contents to different learners in different order depending on the learners' profiles and the
progress they make.
3.3.2 Dynamic Binding
LO repositories established at the sites of content authors who created LOs may be
updated frequently, that is, new LOs are created and existing LOs are enhanced with up-
to-date contents or infrequently used LOs can be removed from the repositories. One way
to leverage LOs stored in the distributed repositories is to register the meta-data and the
access methods of the LOs in a Constraint-based LO Broker that is externally accessible.
Then learners and the Learning Process Execution Engine can query the LO Broker to
find and execute the desired LOs. Another benefit of making LOs programmatically
accessible through the LO Broker is to enhance the effectiveness of learning by selecting
LOs that suit learners' different competencies and backgrounds. The LO Broker works as
a meta-data repository that provides the filtering service.
Dynamic binding refers to the runtime binding of an activity to an LO. It is a more
active approach than altering learning paths in that it can choose different learning
materials dynamically at runtime. It is achieved by allowing a leaf-activity to specify
binding constraints at build-time, and generating a query using the binding constraints to
the LO Broker at runtime. The binding constraints describe the requirements of an LO
that can satisfy the learning objective of a leaf-activity. They are specified in terms of
attribute and inter-attribute constraints which are in the same format as the meta-data of
LOs. The Broker would match the query requirements against the meta-data of the
registered LOs to find the suitable LO(s). If more than one LO match with the query, the
meta-data of these LOs will be presented to the learner so that he/she can choose an LO
of his/her preference. Details about LO registration and binding process will be discussed
in Chapter 5.
In order to make dynamic binding adaptive, binding constraints may contain a
variable in its attribute constraint specification. This variable will be assigned a value
based on a learner's profile, status of objectives, and activity data. In the binding
constraints of a leaf activity, the value on the right-hand side of an attribute constraint can
be prefixed by 'OBJ:' or 'VAR:', which represents the objective, or the activity data,
respectively. At runtime, an objective id prefixed with 'OBJ:' will be replaced by the
status of the objective, and a data id with the prefix 'VAR:' will be substituted by the
value of the activity data at the time of dynamic binding to generate the query. For
example, a binding constraint 'Targetage = VAR:iAge' will be transformed into
'Targetage = 15' when a 15 year old learner is taking the CLO with the activity data
iAge assigned to the age of the learner. As another example, a learner with a disability
can be presented with appropriate content by an attribute constraint 'Media type
ENUMERATION {VAR:content type}', where the activity data 'content type' can be
replaced by an enumeration of strings {"Movie i/ ith caption ", "HTML ", "Flash "} for a
hearing-impaired learner, or {"Audio", "Movie i/ ith audio"} for a visually impaired
learner. Binding constraints allow the discovery of LOs that have appropriate media types.
3.3.3 Model-based Assessment
Assessment is an important aspect of learning in that it measures a learner's
knowledge and problem-solving skills. Assessment results provide important information
that can be used to select learning paths, bind activities to suitable LOs, and identify areas
that a learner should pay special attention to. SCORM's sequencing definition model
does not specify assessment related information, such as when and how an assessment
will be conducted. The content of the presented assessment is solely dependent on an
assessment LO referenced in a leaf activity. Thus, a change of assessment coverage and
location in an activity tree requires a structural change of a learning process and the
authoring of an LO that conducts the new assessment. This would require a substantial
amount of time and effort.
In this work, we allow the specifications of a pre-assessment and a post-assessment
when an activity is defined. We also allow the specification of sequencing rules that can
make use of the results of the assessments to select a learning path, and customize a
learning process structure as discussed before.
Any activity node in a learning process model can carry out a pre- and/or post-
assessment task on the learning contents covered by the tree that is rooted at the node.
The assessment items used in the assessment task can be randomly and proportionally
selected from the assessment items of the non-leaf nodes and those learning objects
bound to the leaf nodes of the tree. The above provision addresses the SCORM's
limitation that only leaf nodes of an activity can have assessments, and gives a content
producer greater flexibility in specifying when and where assessments are to be carried
out.
3.3.3.1 Assessment specification
Assessment specification, which is a part of activity specification, includes
information related to the assessments) to be conducted in an activity. Assessment
Marker describes the type of the assessment (pre- and/or post-assessment) to be
conducted. Assessment Parameters specify the selection percentage (0 100%) of
assessment items and a scoring weight (0 1.0) for each activity that belongs to the
coverage of the assessment. The selection percentage and the scoring weight represent
the contribution of the activity for selecting problems and grading the assessment,
respectively. The number of assessment items that should be extracted from the items in
the activity can be calculated by considering the percentage and the number of
assessment items in the activity. For example, if four pre-assessment items exist in an
activity and the selection percentage for the activity is set to 50%, then two items are
selected randomly from either the activity (in case of non-leaf activity) or the LO bound
to the activity (in case of leaf activity) for the assessment. The scoring weight is applied
to calculate how many points are contributed to the total score of the assessment. The
maximum score, gained score and normalized result score of an assessment are calculated
as follows:
Max Score= Z Pts(P.)* weight(Ai)
A activities in subtree. P1 selected assessment items in A,
, where Pts(P,) is points allocated to P,, and i' eig'ht(4) is scoring weight of A,
Gained Score = PtsGained(P) weight(A,)
4 eactivitiesin subtree. P selected assessmentitems in A,
, where PtsGained(P,) is points gained from P,, and 1i eig/ht(4) is scoring weight of A,.
Normalized result score = Gained Score / Max Score
The normalized result score is set to the measure of the objective associated with
the assessment that has been carried out.
3.3.3.2 Model-based objective measure assignment
As discussed, the satisfaction status of an activity, which is evaluated for the
processing of a roll-up rule of its parent activity, is represented by the status of the
primary objective of the activity. Additionally, non-primary objectives) can be declared
in an activity. A possible usage of non-primary objectives is for the analysis of an
assessment result. In case there is an assessment in an activity, the primary objective's
measure may represent the overall gained score, and the measures of the non-primary
objectives can represent scores gained from some parts of the assessment, that is,
problems chosen from a subset of activities in the assessment coverage. As a result,
objectives can be used in an analysis of the assessment, the result of which can be useful
for giving advice to learners for future study, and making the execution of the learning
process adaptive.
As a part of assessment specification, we provide a mechanism to set multiple
objective measures according to the result of an assessment conducted in an activity. An
objective is paired with a list of activities that belong to the subtree rooted at the activity.
As an example, Figure 3-6 shows a subtree that has an assessment at its root Al. The
activity Al has three objectives: Overall, Left and Right. The objective Overall covers the
whole subtree, Left covers the left subtree, and Right covers the right subtree. A list of
activities (A2, A4, A5, and A6) is assigned to Left, and (A3, A7, A8) to Right. With this
specification, the normalized measures of the objectives Overall, Left and Right will be
set by calculating the maximum score, gained score and normalized score as discussed.
Objectives in Al:
Left
Right
Overall
A1
I I
I I
J]-------- ----L---- -------:
A2 A3
A4 A5 A6 A7 A8
I I I
- - - I - -
LL-
Figure 3-6. Assignment of Measures to Multiple Objectives for an Assessment at Al
Just like there can be multiple quizzes and exams in a course, it is possible to
schedule multiple assessments in a CLO. In addition, it is possible that an assessment is
conducted optionally by a sequencing rule. As an example, Figure 3-7 shows an activity
tree with several assessments. A set of assessment questions will be given at the end of
each of the activities at level 2 under the activity A2, that is, A4, A5, and A6. Only those
learners, who pass less than two of the three assessments, need to take the assessment on
the activity A2. To make the assessment on A2 optional, a roll-up rule of the activity A2
can be specified as "if[At Least Count (2) are Satisfied], then [Satisfy this activity], else
[Unsatisfy this activity]", and a after-post-assessment rule "if [this activity is not
Satisfied], then [perform a Post-Assessment]". Assessment parameters need to be
specified in A2 for the optional assessment.
In the child activities of A3, there will be no assessment, and there will be a
mandatory comprehensive assessment for the entire tree rooted at Al. The roll-up rule of
the non-leaf activity A3 is "If [All child activities are Satisfied], then [Satisfy this
activity], else [Unsatisfy this activity]". Activities without any assessment will be
satisfied after they have been navigated by an after-post-assessment rule "If[true], then
[Satisfy this activity]". The root activity Al has a primary objective primary objective
and a non-primary objective assessment objective for the post-assessment conducted at
the root activity. The entire CLO will be satisfied only if activity A2 and the objective
assessment objective are satisfied, which can be specified by an after-post-assessment
rule of Al: "if[(activity A2 is Satisfied) AND (objective(assessment objective) is
Satisfied)], then [Satisfy this activity], else [Unsatisfy this activity] ".
Level 0 ............ --...... ...
Level 1 ------------- A3
Level 2 --- A4 A5 A6
] Activity without assessment
Level 3 --- Activity with assessment
SActivity With optional assessment
Figure 3-7. Activity Tree with Multiple Assessments
CHAPTER 4
LEARNING CONTENT MODELING LANGUAGE AND LEARNING PROCESS
MODELING LANGUAGE
The learning object models, introduced in Chapter 3, are used for defining ALOs
and CLOs. They are used as the underlying models for the development of authoring
tools to be used by authors of LOs to create ALOs and CLOs. In addition to the
underlying models and authoring tools, we need concrete languages that describe these
two types of LOs without ambiguity. For the purpose of sharing and exchanging the
specifications of ALOs and CLOs, textual languages for defining these two types of LOs
are useful. In this work, a Learning Content Definition Language (LCDL) and a Learning
Process Definition Language (LPDL) were designed using the XML technology for
defining ALOs and CLOs, respectively. Content authors would utilize authoring tools (to
be described in Chapter 5) to define ALOs and CLOs. The tools generate the language
descriptions of ALOs and CLOs using these two languages.
4.1 Why XML?
XML (eXtended Markup Language) has become a de facto standard for
representing structured and semi-structured information in a textual form [Graham, et. al.,
2002]. There are several advantages of the XML technology, which are summarized in
[Software AG, 2004] as follows: "
* XML is a meta language. It defines markup languages for specific purposes and
schemas for all kinds of data models.
* XML is text-based and easy to read.
* XML is ideal for structured data.
* XML is presentation neutral.
* XML is multi-lingual.
* XML is open technology."
Due to the above advantages and the matured XML technology supported by
numerous vendors, XML documents are used across industry and research institutions for
describing new technologies and data exchange.
Storing, parsing, and validating data in an XML document requires the definition of
an XML data schema. Currently there are two ways to define a XML data schema: XML
Schema [W3C, 2001a] and Document Type Definition (DTD). DTD, an older technology,
defines a valid XML document format with its proprietary syntax to define XML
elements and attributes. XML Schema, developed afterwards, supports new features not
supported in DTD, such as definition of new data types, object-oriented concepts
including extension/restriction of base data types and type reuse, and namespace that
enables incorporation of existing XML Schema definitions to create a new one. In
addition, an XML Schema definition is written in a standard XML document to define an
XML data format, which is easier to read than a DTD. Due to these advantages, the
languages and protocols that describe Web-Service technology, such as Web Service
Definition Language (WSDL) [W3C, 2001b], Universal Description, Discovery and
Integration (UDDI) [OASIS, 2004], and Simple Object Access Protocol (SOAP) [W3C,
2003], are defined using XML Schemas. We define the learning object languages, that is,
LCDL and LPDL, using the XML Schema technology. The XML Schema definitions of
LCDL and the LPDL are listed in Appendix A and Appendix B, respectively.
4.2 Meta-model
A meta-model of a modeling language reveals modeling entities, their attributes,
and cardinality and containment relationships among the entities. Figure 4-1 shows the
meta-model diagram of LPDL and LCDL using the class diagram of the Unified
Modeling Language (UML). Each entity, represented by a rectangle in the diagram, is
either an XML element or attribute in the XML-based languages.
I-----------------------------------------------------------
Service Inter-attr LPDL
Constraint Constraint UDDI Saved
links to Registration Time
1 1 1 1
/ I-, --- 1 11 11
II_ Constraints1
o0 1 1 Learning Process
0,1 Business | 1 (CLO) Definition 1
0 1 Business
Service Key binds to 1
S Activity Activity Item on ctor Edge
Binding o r-----
Information Activity l
binds to 1 1 l T 1 1 *1 i
1 1 1 1 1
Updated Limit Objective Activity Activity Assessmentnt Sequencing
Time Data Action Marker Parameters Rules
SContent LCDL
Item
1 Learning Content
+1 (ALO) Definition Practice
1 Item
Assessment
UDDI 1 Learning Items Item
Registration 1I Sequence
Figure 4-1. Meta-model of Learning Process Definition Language and Learning Content
Definition Language and their Relationships
The diagram shows the high-level entities of LPDL and LCDL, and relationships
between the entities, such as cardinality, composition, link and binding. To simplify the
diagram, a super-class 'Activity' is introduced to show that 'Leaf Activity' and 'Non-leaf
Activity' share entities associated with 'Activity'. Details about the meta-model are
explained in the following sections.
4.3 Learning Content Definition Language
LCDL is an XML-based language for defining ALOs. A Learning Content
Definition, that is, an ALO, consists of Content Item, Practice Item, Assessment Item,
Constraints, UDDI Registration, and Learning Items Sequence. A content item contains a
learning asset in the form of text, URL, or HTML that presents the content item. A
practice item and an assessment item include a content item for its problem content, and
an answer. In addition, an assessment item involves a numeric value, that is, a point
allocated to the problem, and a Boolean value that indicates whether the item can be
automatically graded by simple text matching. The following XML code segment
declares learning items consists of four content items and two practice items whose
content type is URL, and two assessment items that have text for their problem
presentation:
http://128.227.176.41:8888/VAM/Lowpressuresystem Ql.
htm
Number="2">
http://128.227.176.41:8888/VAM/Low pressure system lea
rning objectives.htm
http://128.227.176.41:8888/VAM/Low pressuresystem Q2.
htm
http://128.227.176.41:8888/VAM/Low pressure system Ql
Objectives.htm
http://128.227.176.41:8888/VAM/Lowpressuresystem
Ql VAM simulation.htm
40
N/A
http://128.227.176.41:8888/VAM/Lowpressuresystem
Q2 VAM simulation.htm
N/A
Q. If a vaporizer concentration dial is
adjusted to a non-zero setting, will 02 flow generated
by an 02 flush flow through the vaporizer?
A.
No
B. Yes
C. Depends on the setting
of the 02 flowmeter knob
D. Depends on the FGF
setting
A
Q. With the N20 flow meter at its maximum
setting (> 5 L/min), will a hypoxic gas mixture result
at the common gas outlet if the 02 flow meter is
subsequently decreased to its minimum setting (about
200 mL/min)?
A. Yes
B. No
C.
Maybe
B
.......
Learning items sequence describes a presentation sequence of content and practice
items when an ALO is bound to a CLO instance and delivered to learners. Type attribute
of LearningltemSequence element represents the type of the learning item to be presented,
and Itemld attribute matches with Number attribute of the corresponding learning items.
UDDIRegistration records a business interface key and a business service key returned
from a UDDI registry when registering the ALO with the Constraint-based Broker. The
following defines a learning items sequence and the UDDI registration of an ALO:
41
......
Constraints entity is the meta-data that describes an ALO, and is used for the
registration and discovery of the ALO through the Constraint-based Broker. The
following is an example attribute constraint specification describing that 'Title' is "Low
Pressure System VAM", 'MediaFormat' is "HTML" and "Shockwave", 'Language' is
"English", 'cost' is 0, and 'Author' is "Gilliean Lee":
Title
string
eq
"Low Pressure System VAM"
no
0
Media Format
string
enu
{"HTML", "Shockwave"}
no
0
Language
string
eq
"English"
no
0
Cost
float
eq
0.0
no
0
Author
string
eq
"Gilliean Lee"
no
0
.......
The complete definition of the ALO shown in the above example can be found in
Appendix C. Figure 4-2 shows the detailed meta-model of LCDL, consisting of entities
and relationships among the entities.
1 o0..1 =Text|
Content 1 o..1
1 1.. Item 1 URL
I Learning Content A1 It l m
IConstraints (ALO) Definition
1 1
SPractice Answer
S Item
Service Inter-attr
Constraint Constraint 1 1 1
[_ UDDI.I 1i I i "IIAssessment Point
UDDI Learning Items Item
Registration Sequence Autograde
Figure 4-2. Detailed Meta-model for LCDL
4.4 Learning Process Definition Language
LPDL is an XML-based language used for defining a CLO. LPDL specifies a
learning process in the form of an activity tree. As discussed in Chapter 3, there are
several modeling constructs of the activity tree: leaf activity, non-leaf activity, connector,
and edge, each of which is expressed as an XML element in LPDL. The most
fundamental modeling entities are LeafActivity and Non-leafActivity, with which the
other entities such as Limit, Objective, Activity Data, Activity Action, Binding Info,
Assessment Marker, Assessment Parameters, ETR Rules, Content items, Practice items,
and Assessment items are associated. The related entities will be explained when the
Activity entities are introduced.
In addition to the modeling constructs of the activity tree, Constraints and UDDI
Registration entities are also specified. Saved Time entity stores the most recent time a
CLO is created and saved. The information is compared with a deployment time recorded
in the LPEE to verify whether a re-deployment of the CLO is necessary or not.
We will use a CLO, which models the instructional process of a Virtual Anesthesia
Machine, as an example of LPDL code in this section. The full specification of the CLO
can be found in Appendix C.
4.4.1 Non-leaf Activity
In an activity tree, a non-leaf activity is a node that has at least one child node. It
can contain contextual contents, that is, introduction and summary, and practices and
assessments for exercising and assessing a learner's ability to integrate the contents
presented in its child activities. The following XML code segment specifies that an
introduction can be accessed through the given URL and a summary is given in the text.
http://www.vam.org/VAM/VAM Intro.htm
Original Virtual Anesthesia Machine Web page is
http://vam.anest.ufl.edu
.
Special Thanks to Dr. Sem Lampotang
Copyright: University of Florida.
.......
The Limit entity specifies a delivery limit of the activity in terms of the time
experienced or the number of attempts. The Objective entity represents a learning
objective of the activity. An activity is required to have a primary objective, which
represents the satisfaction status of the activity, as discussed in Chapter 3. The Activity
Data entity contains attributes or variables of simple data types supported by the Java
programming language. Their values specify a learner's profile information, the result of
some simple calculations, or the current status of an activity. They can be used in the
processing of a CLO instance. At runtime, all the values of Activity Data are passed to
sequencing condition-action rules, such as Pre-Activity, After-Pre-Assessment, and
After-Post-Assessment rules, for communication between the CLO runtime instance and
the rule code. The following XML code segment declares the objectives and activity data
of an activity named 'Part_3_SafetyExercises':
SatisfiedByMeasure="true"
MinimumSatisfiedNormalizedMeasure="0.5" />
SatisfiedByMeasure="true"
MinimumSatisfiedNormalizedMeasure="0.5" />
SatisfiedByMeasure="true"
MinimumSatisfiedNormalizedMeasure="0.5" />
Id="oBreathingCircuit" SatisfiedByMeasure="true"
MinimumSatisfiedNormalizedMeasure="0.5" />
Id="oScavengingSystem" SatisfiedByMeasure="true"
MinimumSatisfiedNormalizedMeasure="0.5" />
.......
The Activity action, a piece of Java code executed at a specific stage of activity
processing, can be defined to perform some simple operations such as checking learner
profile and status of objectives, assigning values to activity data, carrying out a data type
conversion, and so on. In a non-leaf activity, there are 'Before-Pre-Activity Actions',
'Before-Pre-Assessment Actions', 'Before-Introduction Actions', 'After-Introduction
Actions', 'Before-Rollup Actions', 'Before-Post-Assessment Actions', and 'After-Post-
Assessment Actions'. Each action is translated into Java code.
Internally, an activity code class is created for each activity. In the class, activity
data are declared as member variables, and a method implements each of the activity
actions. LPEE invokes these methods at specific stages of activity processing. Table 4-1
summarizes activity actions in a non-leaf activity and their descriptions.
Table 4-1 Execution Stages and Description of Activity Actions in Non-leaf Activity
Activity Action Name Execution Stage Description
Before-Pre-Activity Before processing Pre-Activity Assign values to activity data to be
rules used for the execution of Pre-
Activity rules.
Before-Pre-Assessment Before conducting a pre- Executed to take any necessary
assessment action for the preparation of a pre-
assessment.
Before-Introduction Before processing After-Pre- Assign values to activity data to be
Assessment rules, i.e. before used for the execution of After-Pre-
presenting a content item as the Assessment rules.
introduction
After- Introduction Before drill-down to a child activity, Assign values to activity data to be
i.e. before processing Drill-down used for the execution of Drill-down
rules rules.
Before-Rollup Before processing Roll-up rules, Assign values to activity data to be
i.e., before roll-up to itself after used for the execution of Roll-up
finishing a child activity rules.
Before-Post- Before conducting a post- Executed to take any necessary
Assessment assessment action for the preparation of post-
assessment.
After-Post-Assessment Before processing After-Post- Assign values to activity data to be
Assessment rules used for the execution of After-
Post-Assessment rules.
Assessment information for the model-based assessment discussed in Chapter 3 is
specified as a part of an activity. The Assessment Marker specifies whether a pre- and/or
post-assessment is going to be conducted in the activity. For each assessment, there exist
the Assessment Parameters, which include the Assessment Selection Parameters and the
Objective Settings. The Assessment Selection Parameters specify the activities
participating in the assessment. It is composed of a set of tuples (activity ID, selection
percentage, scoring weight), where activity ID is an identification of a descendent
activity from which assessment items will be collected. Selection percentage specifies
what percentage of assessment items will be chosen from the activity, and scoring weight
is the weight of selected assessment items to be used for grading the assessment.
46
Objective Setting entity, which is a subset of descendent activities, indicates the activities
that contribute to the calculation of the objective's measure.
The following XML code segment shows an example of assessment parameters
specified in a parent activity 'Part 3_Safety Exercises', which has the following child
activities: 'HighPressure_System', 'Low_Pressure_System', 'Breathing_Circuit',
'ManualVentilation', 'MechanicalVentilation', and 'Scavenging_System'. In the
parent activity, there are objectives 'oOverall', 'oPressureSystem', 'oBreathingCircuit',
'oVentilation', and 'oScavengingSystem' whose measures will be assigned as the result
of an assessment conducted in the activity:
ScoringWeight="l" SelectionPercentage="0.5" />
ScoringWeight="l" SelectionPercentage="0.5" />
ScoringWeight="l" SelectionPercentage="0.5" />
ScoringWeight="l" SelectionPercentage="0.5" />
ScoringWeight="l" SelectionPercentage="0.5" />
ScoringWeight="l" SelectionPercentage="0.5" />
ScoringWeight="l" SelectionPercentage="0.5" />
Part 3 Safety Exercises
High Pressure System
Low Pressure System
BreathingCircuit
ManualVentilation
MechanicalVentilation
Scavenging_System
High Pressure System
Low Pressure System
BreathingCircuit
Manual Ventilation
Mechanical Ventilation
Scavenging_System
......
Minimum 1
Satisfied I
Measure I
Figure 4-3. Entities in the Meta-model of a Non-leaf Activity in Learning Process
Definition Language
As discussed, a condition-action rule has the structure "If [Condition], then
[Action], else [Alt-Action]". Thus, the Rule entity consists of a Condition, an Action, and
an Alternative Action. Condition is in the form of a Boolean expression, and Action and
Alternative Action are Java code segments to be executed when the Condition is satisfied
or unsatisfied, respectively. Sequencing rules for a non-leaf activity include 'Pre-activity',
'After-Pre-Assessment', 'Drill-down', 'Roll-up', and 'After-Post-Assessment' rules.
Figure 4-3 depicts a detailed meta-model of a Non-leaf Activity entity with its related
entities.
4.4.2 Leaf Activity
A leaf activity is a leaf node in an activity tree. Unlike a non-leaf activity, which
includes learning items in it, a leaf activity binds to an LO at runtime either statically or
dynamically. Binding Info specifies the method of LO binding and the required
information for each method. In the case of static binding, a business service key and a
business interface key is specified. Dynamic binding involves the requirements of LO
that the activity needs to bind, which are specified in the same format as the meta-data of
LOs. Implementation details on the registration and dynamic binding mechanism will be
given in Chapter 5. In the following example of Binding Info, it specifies the information
necessary for dynamic binding, where the attribute Keywords should be "input output of
anesthesia machine" and the attribute Language should be set to the value of activity data
'sLanguage' at the time of binding:
Keywords
string
eq
"input output of anesthesia machine"
No
0
Language
string
eq
"VAR:sLanguage"
No
0
.......
Table 4-2 Execution Stages and Description of Activity Actions in Leaf Activity
Activity Action Name Execution Stage Description
Before-Pre-Activity Before processing Pre-Activity Assign values to activity data to be
rules used for the execution of Pre-Activity
rules.
Before-Binding Before binding to a learning Assign values to activity data that can
object be referenced for updating dynamic
binding constraints.
Before-Pre-Assessment Before conducting a pre- Executed to take any necessary
assessment action for the preparation of pre-
assessment.
Before-Content Before processing After-Pre- Assign values to activity data to be
Assessment rules, i.e. before used for the execution of After-Pre-
presenting Introduction content Assessment rules.
item
Before-Post- Before conducting a post- Executed to take any necessary
Assessment assessment action for the preparation of post-
assessment.
After-Post-Assessment Before processing After-Post- Assign values to activity data to be
Assessment rules used for the execution of After-Post-
Assessment rules.
In a leaf activity, there are the following activity actions: 'Before-Pre-Activity
Actions', 'Before-Binding Actions', 'Before-Pre-Assessment Actions', 'Before-Content
Actions', 'Before-Post-Assessment Actions', and 'After-Post-Assessment Actions'.
Table 4-2 describes the execution stages and the purposes of the activity actions. A leaf
activity specifies Pre-activity, After-Pre-Assessment, and After-Post-Assessment
sequencing rules.
Figure 4-4. Meta-model of Leaf Activity and Related Entities in the Learning Process
Definition Language
The definitions of the Limit, Activity Data, Objective, Assessment Marker,
Assessment Parameters entities in Leaf Activities are identical to those in Non-leaf
Activities. Figure 4-4 shows the detailed meta-model of Leaf Activity and its related
entities.
4.4.3 Connector
A connector relates a parent activity and its child activities in a cluster. Sequencing
ControlModes of the cluster are specified in a connector. Additionally, it can include an
ordered list of child activities, which gives the processing sequence of these child
51
activities in the Forward Only and Flow sequencing control modes. Figure 4-5 shows the
detailed meta-model for connector, edge, and their related entities.
Sequencing 1 1 1 1
Control Modes Connector Edge
1 1 1 1..2
0..1 0..1 ..1 1 ..*
Choice Flow Forward 1
Only Activity
Mode Mode Mode (Leaf, Non-leaf)
Figure 4-5 Meta-model of Connector and Edge in the Learning Process Definition
Language
The following XML code defines two connectors having the 'ForwardOnly' and
'Flow' sequencing control modes, respectively:
Part 1 Basic Concepts
Part 2 IO of VAM
Part 3 Safety_Exercises
HighPressure System
Low Pressure System
Breathing Circuit
Manual Ventilation
Mechanical Ventilation
Scavenging System
.......
4.4.4 Edge
An Edge is a directed edge, which connects an activity to a connector or a
connector to an activity. It has a 'from' attribute and a 'to' attribute. The value of each
attribute can be either an activity ID or a connector ID. The meta-model of Edge is shown
in Figure 4-5. The next XML code defines edges that connect an activity to a connector,
and vice versa:
Id="connectorl::Part 1 BasicConcepts" From="connectorl" />
From="connectorl" />
Id="connectorl::Part 3 Safety Exercises" From="connectorl" />
52
From="Part 3 Safety_Exercises" />
Id="connector2::High PressureSystem" From="connector2" />
From="connector2" />
From="connector2" />
From="connector2" />
Id="connector2::Mechanical Ventilation" From="connector2" />
From="connector2" /> ......
CHAPTER 5
E-LEARNING SERVICE INFRASTRUCTURE
We envision a new type of virtual e-learning community that can be established
based on reusable, assessable, and programmatically accessible e-learning objects that are
created and stored in a distributed fashion over the Internet.
5.1 E-Learning Infrastructure for Virtual E-Learning Communities
The e-learning infrastructure provides a dynamic and collaborative e-leaming
environment in which a large number of virtual collaborative e-learning communities can
be dynamically established and evolved. Each collaborative community is dynamically
formed by a community of Internet users who have a common interest in a specific
subject of learning, and who can provide and/or use the knowledge for solving problems
in that subject domain.
5.1.1 Roles in a Virtual E-learning Community
Members of a virtual e-learning community can play the following roles: content
provider, content composer, content evaluator, content learner, and community host
administrator. A member can play one or more of these roles.
Content providers create and register ALOs. Since an ALO is an aggregation of learning
items (i.e., content, practice and assessment items), which make references to learning
assets, a content provider can either take advantage of the learning assets available in the
Internet, or produce learning assets by using appropriate authoring tools to build ALOs.
HTML authoring tools, audio and movie editing tools, Macromedia Director, and so on,
can be used to create learning assets. The constructed ALOs are stored in the provider's
local repository, maintained by the Apache Xindice XML DBMS. Each ALO, together
with its meta-data and business service key, is registered as a Web-service in the
Constraints-based LO Broker deployed at a community host's site for later discovery and
delivery. Community members, and the Learning Process Management System can
browse and query the LO Broker for LOs. The LO Broker performs constraints matching
to discover the best LO that matches with a learner's request and profile.
Content composers' responsibility is to compose and register CLOs by making use
of the CLO authoring tool. Because non-leaf activities can include learning items, they
also need to be able to access learning assets as content providers. They can work with
content providers, or take the role of a content provider. Constructed CLOs are stored in
the content composer's local repository. The meta-data and access information of the
CLO is again registered with the LO Broker and made accessible and reusable to other
community members.
Content evaluators, elected by a virtual e-leaming community, would review,
evaluate and rate the ALOs and CLOs that are registered with the LO Broker based on
learners' feedback and LOs' use statistics, and, together with the administrator, approve
or disapprove their inclusion for the community's use, and their modification and
removal. It is necessary that content providers, content composers, and content evaluators
are domain experts in the subject area of LOs that they produce and evaluate.
Administrators of the community host are responsible for the operation and
maintenance of the Learning Process Management System installed at the host site.
Additionally, they take care of the registration and enforcement of the community's
policy rules, which govern LO registrations with the LO Broker. Policy rules can
recommend to members the LOs that have high evaluation scores, or drop LOs from the
registry that are rarely used or have very low rating.
Content
Evaluators
rUses
Content G e te Ev ua by c e Evaluation Tool
Providers Content va ted
Composers
Generate
aI LOs Used by Used b CLOs
Content Re i ed '
Providers d2 s Leev aluq n LOs
ProvidesElearning Host
Pf.S P providess E-learning Adminis
Content Register Profile Commun [ a
...Content ---- CCommunity intain E ab/sh
Learners Host LO action
Community Rules
Figure 5-1. Roles and Relationships among Members of a Virtual E-Leaming
Community
Content learners are members of a virtual e-learning community who use the LOs
created by the virtual community. During the learner registration, they are required to
provide information that constitute learner profiles, which are used by the Learning
Process Management System to achieve adaptive and customizable delivery of learning
experience. In addition to learning from LOs, they are authorized to evaluate the LOs that
have been delivered to them. They make use of the Web User Interface provided by the
community host. The roles and relationships among members of a virtual e-leaming
community are depicted in Figure 5-1.
5.1.2 Web-Services-Based E-Learning Service Infrastructure
The infrastructure is composed of distributed software components including the
LO authoring tools, the distributed LO repositories, the Learning Process Management
System, the LO Broker, and Web Servers. Each network site of a virtual community has a
Web server to provide HTTP and SOAP Web-services. LO repositories are stored at the
sites where they are created to avoid the communication/storage bottleneck of a
centralized repository and to give more authority to content providers and content
composers in their creation, modification and removal of LOs.
Different software components are installed at different sites depending on the roles
that the members of a virtual community play. The Content Provider site is equipped
with the ALO authoring tool and other learning asset authoring tools. It also has an XML
DBMS to store and maintain the local LO repository. The Content Composer site has a
CLO authoring tool in addition to the facilities available to a content provider. Using the
LO authoring tools, content composers and providers store LOs in their local LO
repositories, and register them with the Constraints-based LO Broker installed at the host
site. The Community Host site has the Learning Process Management System, the
Constraints-based LO Broker and a local LO repository. The local LO repository
maintains all the data necessary for the execution and delivery of LOs to learners, that is,
the learner profile models, the runtime status models and LOs that are fetched from the
distributed repositories at the sites of content providers and composers for processing by
the Learning Process Management System. The Content Learner site needs only a Web
browser. A learner can browse or query the LO Broker for registered LOs by specifying
his/her learning requirements in terms of the meta-data and constraints associated with a
desired LO. The learner can then store the LOs of his/her choice in a learner space, called
MyCLOs, for later execution. He/she can execute the LOs in his/her learner space. The
facilities at different sites of the web-service-based e-learning service infrastructure are
shown in Figure 5-2. The implementation details of the infrastructure are discussed in the
next section.
egistry Content Provider
Register, Lookup LO Access Info ALO
Community Host Authoring Tool Web-page -
authoring tool I
Link A sets
Broker (UPL)
Register Import
ALOsHTMs
Learning .
Process LLnk assets Learning
Management rnel (URL) Asset
System C Create i
Web, SOAP Server Web, SOAP Server
Request ALOs
Request CLOs
Register CLOs | -
Search LOs, Web Servers
Request LO Execution, Content Composer V"b, SOAP on Internet
Content Delivery Create, Edit CLOs I Server
Web earning
Browser Authorng Tool Assets
Content Learner [
Figure 5-2. Web-service-based E-Learning Service Infrastructure
5.2 Implementation
E-leaming services enable the authoring, registration, delivery and management of
LOs. They are needed for the creation of virtual e-learning communities. In order to
provide these e-leaming services, a service framework that facilitates the authoring,
registration, discovery, binding, and invocation of LOs is necessary. In addition, a
uniform way of modeling multimedia learning assets and Web-based service components
in the form of objects is needed. The Web-services technology developed by the IT
industry can be adapted and extended to serve this purpose.
5.2.1 Constraints-enabled Web-service Model
The Universal Description, Discovery and Integration (UDDI) [OASIS, 2004]
provides a general framework to allow objects to be defined as Web-services using the
Web Service Description Language (WSDL). The major elements of WSDL used for
Web service description are as follows [W3C, 2001]. The relationships among these
elements are illustrated in Figure 5-3:
* types a container for data type definitions using some type system (such as XSD).
* message an abstract, typed definition of the data being communicated. It consists
of the data types defined in 'types' elements.
* operation an abstract description of an action supported by the service. Input /
output parameters are defined using the messages defined in 'message' elements.
* portType an abstract set of operations supported by one or more endpoints.
* binding a concrete protocol and data format specification for a particular port
type.
* port a single endpoint defined as a combination of a binding and a network
address.
* service a collection of related endpoints.
A WSDL document can be posted on the Internet, and its access point (URL)
together with a textual description and the category information can be registered with a
Web-services registry such as the UDDI registry. The contents of the registry can then be
navigated or searched manually or programmatically to discover and obtain the access
information of suitable Web-services. As an alternative to downloading identified objects
for local use, the Web-services model allows the runtime binding of a service request to a
remote Web-service, and provides the Simple Object Access Protocol (SOAP) [W3C,
2003] that allows the activation of a remote service by exchanging SOAP XML messages.
By applying the Web-service technology, LOs and other instructional resources, tools,
and software systems can be uniformly specified as Web-services. However, the existing
specification and implementation of UDDI does not provide constraints specification and
filtering capabilities to match service constraints with a service requestor's requirements
to identify suitable services. Without these constraints specification and filtering
capabilities, an LO can be selected by the registry, but the selected LO may not satisfy
the requestor's requirements.
In this work, we have extended the WSDL specification and the UDDI registry's
capabilities by allowing constraints to be specified in WSDL, and adding constraint
processing capability to the UDDI registry. In the extended WSDL, a constraints element
that describes the meta-data of a Web-service is written using the LO meta-data format,
and appended as a part of the Web-service description. The extended WSDL information
model is shown in Figure 5-3.
par t,p Service Interface Definition
I. It portType
(abstract)
message (abstract)
operation
I concrete implementation
binding
(concrete)
message (concrete)
operation
service + T Concrete en'.:o:.lni pron
4- Made concrete by
Contains zero or more C:onsraln
Service Implementation Definition
Figure 5-3. Extended WSDL Information Model (In part from [Graham, et. al. 2002])
Figure 5-4 shows a part of an extended WSDL of an LO that includes an operation
getLO, the access point of which is http://128.227.176.41:8888/axis/LODBService.jws.
The meta-data attribute 'Keywords' of the LO has the enumerated values: "Anesthesia
Machine", "Anesthesia", "Virtual Anesthesia Machine", "Simulation", and "Low
Pressure System", the attribute 'Media Format' has the enumerated values "HTML" and
"Shockwave", and the attributed 'Cost' has 0 as its value.
xmlns="http://schemas.xmlsoap.org/wsdl/"
xmlns:apachesoap="http://xml.apache.org/xml-soap" xmlns:impl="urn:LODBService"
xmlns:intf="urn:LODBService"
xmlns:soapenc="http://schemas.xmlsoap.org/soap/encoding/"
xmlns:wsdl="http://schemas.xmlsoap.org/wsdl/"
xmlns:wsdlsoap="http://schemas.xmlsoap.org/wsdl/soap/"
xmlns:xsd="http://www.w3.org/2001/XMLSchema">
transport="http://schemas.xmlsoap.org/soap/http" />
encodingStyle="http://schemas.xmlsoap.org/soap/encoding/"
namespace="urn:LODBService" use="encoded" />
encodingStyle="http://schemas.xmlsoap.org/soap/encoding/"
namespace="urn:LODBService" use="encoded" />
location="http://128.227.176.41:8888/axis/LODBService.jws" />
Keywords
string
enu
{"Anesthesia Machine", "Anesthesia", "Virtual Anesthesia
Machine", "Simulation", "Low Pressure System", "High Pressure
System", "Breathing Circuit", "Manual Ventilation", "Mechanical
Ventilation", "Scavenging System"}
No
0
Media_Format
string
enu
{"HTML", "Shockwave"}
No
0
Cost
float
eq
O.0
No
0
Figure 5-4. Extended WSDL Document Example
A Constraints-based Web service Registry [Degwekar, Su & Lam, 2004], which
stores constraints and performs constraints matching to select suitable Web-services, is
utilized as a constraints-enabled Web-services registry. The registry leverages Apache
jUDDI registry (UDDI v.2.0) [The Apache Software Foundation, 2003] as a back-end to
acquire globally unique registration keys and store access points of LO repositories where
LOs are located. Since we make use of the registry as an LO registry that matches LOs
with browse requests, we shall refer it as the 'Constraints-based LO Broker'.
LOs registered as Web services are accessible by calling Web-service operations of
the LO repositories (getLO, addLO, etc) deployed at the access points of the Web-
62
services. The registry serves as a meta-data repository in contrast to LO repositories.
More about the Constraints-based Registry will be discussed in a later section.
5.2.2 Authoring Tools & LO Repository
Authoring tools with graphical user interfaces have been implemented to help
authors compose ALOs and CLOs efficiently and effectively. We make use of the Java
Architecture for XML Binding (JAXB) [Sun Microsystems, 2003] technology to
implement XML parsing and updating in our authoring tools. The JAXB package
generates a set of Java class files using an XML Schema definition as an input. The use
of JAXB package and the generated Java class files greatly simplifies the parsing,
updating, writing, and validation of XML files, which conform to the XML Schema
definition provided as an input.
I.- I- .. ,,n
I I I 1 1 I I
"- ,
-I .I .. T*L .2W..! 5! !
,i l I i I IT I ,,w
II '- I f F .- -
A B
Figure 5-5. Graphical User Interface Frame of the ALO Authoring Tool. A) Content
Items Authoring Tab. B) Service Constraints (Meta-data) Tab.
5.2.2.1 ALO authoring tool
We have developed an ALO authoring tool that facilitates authoring and
registration of ALOs. It is developed using Java Swing components. Its user interface is
63
similar to a tabbed dialog box in a windows-based operating system. The tool is used to
add, delete, and edit learning items, and to create, edit and register ALOs. The user
interface of the
chema Process Model Activity
IWl0l@ llE
ALO authoring tool is shown in Figure 5-5.
edit IrporfRemove Code Generation
d Process Model
9 O LeafActiviles
D Part_1_Basic_Conce
D Part_ 2__of_ VAM
D High_Pressure_Syst
[ LowPressureSyste
i BreathingCircuit
Li ManualVentlation
[ Mechanical Ventilatlc
[ scavenging_Syslem
9 L NonleafArtivities
DVAM
e Part 3_Safety Eer
9 [Connectors
connector
connector2
L 1 Edges
SVAM connector
L connector Part1_B
i connectorl Part 2_(
Sconnectorl Part3_S
D Part_3_Safety_Exerc
D[ connector2 High_Pr
D connector Low_re
[ connector Breathanc
D connector Manual
Sconnec tor2 Mecha.
4_:_1 I
-
VAM
S LE
Part 1_BasicConcepts Part_2_10_ VAM Part_3_Safet_Exercises
E
C
IE E I E E
High PressureSystem Low_Pressure_Systern Breathing_Circull Manual Ventllation Me
2I"'
Tree View
ActivtyD Pal _oAM
Activity Name t otoVM
Description simulation works Including Input and Output
Lim it----------------
O Num of Attempts E Times
Assessment Marker
OPreAssessment Set Parameters
0 nems from This Actty Only
0 iems from Subtree
D PostAssessment | et Parameters |
Binding Info for Leaf Acttiv
BrokerURL I
SDynamic Bdinig et Matching Constraints
0 Static Binding
Service Key
Objective ID
loOverall
Activity IDs in Assessment
Part 3 Safety Exercises '
High Pressure System
Low Pressure System
BreathingCircuit
Manual Ventilation -
Add Objective Setting
Registered Objective Setting
Obective ID Associated Activity IDs
oOverall Part_3_Safety_Exercises,High_Pressu
oPressureSystem HighPressure System,Low Pressure
oBreathingCircuit Breathing_Circuit
oVentilation Manual VentilationMechanical Ventilat
Remove Objective Setting
OK ancel B OK Cancel
Figure 5-6. Graphical User Interface Frame of the CLO Authoring Tool. A) User
Interface Frame with Tree View and Canvas Area. B) Leaf Activity Dialog
Box. C) Assessment Dialog Box.
anical ventilation
Scavenging System
Canvas Area
8~0~ II~
-1--------------------------------------------------------
1 111Obiectives -ings
rrmr~me! WI~
)1 in1lnrpl ',pliiJ'
I
I
5.2.2.2 CLO authoring tool
The CLO authoring tool supports the authoring and registration of CLOs. As shown
in Figure 5-6 (A), the tool's frame provides a canvas area on the right side for designing
an activity tree graphically, and a tree view on the left side of the frame where modeling
constructs are categorized and presented. Using the CLO authoring tool, a content
composer can create, edit and register CLOs. Dialog boxes for editing the properties of
the modeling constructs such as activities, connectors and edges are also provided.
Functions such as 'Add', 'Delete' and 'Edit' learning items are provided for defining
learning items in non-leaf activities. Activity and Assessment dialog boxes are shown in
Figure 5-6 (B) and (C).
5.2.2.3 LO repository
The ALO and CLO authoring tools store and retrieve LOs from an LO repository
stored in an XML database managed by the Web-based native XML DBMS, Xindice
[The Apache Software Foundation, 2004]. The Learning Process Management System
also accesses LOs from distributed LO repositories for processing. For interoperability
and ease of access, each LO repository is implemented as a Web-service. The Java Web
Service (JWS) technology supported by the Apache Axis SOAP engine is used for the
deployment of the Web-service of the LO repository. Any part of an LO, such as meta-
data, contents items, assessment items and practice items, can be accessed from the LO
repository. The WSDL document that describes the LO repository's Web-service is given
in Appendix D. Details about the LO registration process with the Constraints-based LO
Broker and the accessing of LOs from LO repositories will be discussed later.
5.2.3 Learning Process Management System
LOs are registered, discovered and processed by the Learning Process Management
System, which consists of the following main components: a Learning Process Execution
Engine (LPEE), an Event-Trigger-Rule (ETR) Server, an Assessment Component, an LO
Broker, a local LO Repository, and a User Interface Component. The architecture of
Learning Process Management System is illustrated in Figure 5-7. The components of the
system are discussed in the following sections.
i------------------------------------------------------------------
Learning Process Management System Communication Protocol
4- Java RMI
---- ---- -------- ------- SOAP
LO Broker ___ UI- u-" Learning Process 4------ HTTP
I ---- Component Execution Engine
Monitoring Component
Component
SI a Collaboration Tools
S LO Repository, (E-mail, Internet
Runtime Data, ETR Messenger,
External Learner ProfilesServer Workflow, etc)
LO Repositories Web-Service
I0I (HTTP, SOAP
Server)
Figure 5-7. Architecture of Learning Process Management System
5.2.3.1 Learning process execution engine
The LPEE processes LOs using the rule-based execution model to facilitate active,
adaptive, flexible and customizable e-leaming service. When a learner requests the
execution of an ALO, a predefined CLO having a single leaf activity is instantiated and
the leaf activity is bound to the ALO. When the execution of a CLO is requested, LPEE
fetches the requested CLO from a repository that stores the CLO by making a Web-
service call, and then deploys the CLO by generating activity codes and installing events,
triggers and rules specified in the CLO in the Event-Trigger-Rule Server. Next, an
instance of CLO is created for execution. A runtime status model, described in an XML
document, is created and maintained in a local LO repository stored in an XML database.
* Runtime Status Model
The model keeps the following information necessary for the execution of a CLO:
o Status of the instance: started, suspended, and finished.
o Current activity id.
o Status of activities: unattempted, attempted, and completed.
o Status of learning objectives: unknown, satisfied, unsatisfied, and current
measure.
o Status of activity attempts.
Upon the creation of a CLO instance, a runtime status model described in an XML
document, and a temporary copy of the CLO are created and maintained in the local LO
repository. The local LO repository is built on top of Xindice XML database, which
provides the persistent storage. LPEE maintains the runtime status model document for
the CLO instance. Upon receiving sequencing commands, the states of activities and the
instance are recorded in the runtime status model according to the state transition
diagrams shown in Figure 5-8. The runtime status model is defined in an XML Schema
shown in Appendix E.
Start Suspend Attempt
StartStart
started I Suspended
Unattempted Attempted
Resume
Finish, Finish Exit
Terminated Attempt
Finished
A Completed
Figure 5-8. State Transition Diagrams for Learning Process and Activity. A) Process
Instance Status Transition Diagram B) Activity Status Transition Diagram
A monitoring component makes use of the runtime status model to provide the
status information about the CLO instance to a learner. A UI component invokes the
monitoring component to create a Web page that presents the status information along
with the activity tree of the CLO instance.
* Architecture of LPEE
A learning process instance handler object (LPInstance Handler) takes care of the
execution of a learning process (CLO) instance by managing the execution of its
activities and maintaining its runtime status model through an Activity Handler and the
LO Repository Access Module. An Activity Handler object is responsible for the
execution of an activity. It instantiates the activity code object of an activity and invokes
its appropriate methods in accordance with the stage of execution. Some of the methods
post events to the ETR Server to invoke sequencing rules that are installed when the CLO
is deployed. The parsed structure of the CLO definition is maintained so that the LO
specification can be retrieved rapidly. The runtime objects of a CLO instance and their
relationships with a learning object definition are illustrated in Figure 5-9.
ETR Server
1 Condition-Action
ETR Server Rules
Rules
Figure 5-9. Relationships between Learning Object Definition and Runtime Objects of a
CLO Instance
68
The LMS Engine Core object implements the core functionality of LPEE. The
runtime objects, including activity codes, parsed CLOs, LPInstance Handlers, and
Activity Handlers, are maintained in four hash tables. Using the CLO instance ID or a
combination the instance ID and an activity ID, the runtime objects of a CLO instance
can be retrieved from the hash tables. LPEE is encapsulated as a Remote Method
Invocation (RMI) server accessible to sequencing rules and the UI component.
LMS Engine Core
Activity Code Hashtable Learning Process Hashtable
Key: LPlnstance Id + Activity Id Key: LPlnstance Id
Activity Code Parsed CLO
LP Instance Hashtable Activity Handler Hashtable
Key: LPlnstance Id Key: LPlnstance Id + Activity Id
LPInstance Handler Activity Handler
Runtime Status Model Learner Info Model Learning Object DB
Accessor Accessor Accessor
LMSEnginelmpl
(RMI Server)
Figure 5-10. Architecture of the Learning Process Execution Engine
The LO Repository Access Module consists of three classes that access the Xindice
XML database: Runtime Status Model Accessor, Learner Information Model Accessor,
and Learning Object DB Accessor. These classes make use of XPath and XUpdate
technologies to query and update the XML database. The Runtime Status ModelAccessor
class is responsible for the management of runtime status models. The Learner
Information Model Accessor class queries and updates the learner profile models. The
Learning Object DB Accessor class stores temporary LOs and accesses LO specifications.
The architecture of LPEE is shown in Figure 5-10.
S LO Deployment Process
A CLO has to be deployed before an instance of the CLO can be instantiated. The
deployment process involves the generation of activity codes for all the activities of the
CLO, and the installation of sequencing rules, which involves code generation for events,
triggers, and rules, and installing the code in the ETR Server. During the deployment
process, the installProcesslnstance method of the LMS Engine Core object invokes the
deployCLO method of CLO RuntimeDB Access object. The deployCLO method checks
whether the CLO has to be deployed by comparing the CLO's last deployment time and
the CLO's update time. If the CLO is deployed for the first time, the deployCLO method
will generate code for events, triggers, rules, and activities that are necessary for the
execution of the CLO instance. If the CLO has been deployed before and the update time
is later than the deployment time, then some of the activities of the CLO have to be
deployed to reflect the update. The deployment method deploys only the activities whose
update time is later than the deployment time by generating and installing codes for the
events, triggers and rules associated with the updated activities and the codes for the
updated activities. An object diagram of the objects related to the deployment process is
shown in Figure 5-11.
Non-leaf Activity
Code Generator
Figure 5-11. Object Diagram of CLO Deployment Process
* Learner Profile Model
Learner profile information provides the basis for the adaptation and customization
of learning objects. An LO can give different learning experiences to different learners
with diverse competencies and backgrounds. The IMS Learner Information Package
Specification [IMS, 2005] provides a specification of learner information model. Our
Learner Profile Model adopts most of the core information in the IMS specification and
adds additional features such as the specification of preferred learning style as shown
below:
o Identification Learner ID, Password, Address, Name, Contact Information
(phone, e-mail, fax, etc.), Date of Birth, Gender, Place of Birth.
o Goal Description, Setup data, Target date, Completion Date, Status (active,
inactive, completed), Sub-goals.
o Qualification, Certificate, License
o Accessibility
Language Proficiency Oral Speak, Oral Comprehension, Read, Write
Preference
Network Connection Dial up, DSL, T1, T2, etc.
Preferred Learning Style Receptive, Directive, Guided Discovery,
Explorative.
Preferred Media Type Movie, Audio, Document, Flash, Presentation,
etc.
Disability Hearing, Visual, Mobility.
o Learning History Start data, Finish date, Evaluation Result, Status (Active,
Suspended, Finished, Failed).
o Affiliation Affiliated organizations and related information.
o Interest Description.
The XML Schema Definition of the learner profile model is shown in Appendix E.
5.2.3.2 Event-Trigger-Rule server
The ETR Server is the rule engine that implements and executes sequencing rules
and community policies and regulations. These rules, policies and regulations are defined
in terms of events, triggers and condition-action rules. In the next few paragraphs, we
explain how a sequencing rule of a CLO is implemented as an event, a trigger and a rule,
and how the sequencing rule is executed.
An event is a data structure that can be posted to represent an occurrence of an
event type, and is implemented as a Java class. The attributes of an event class include
CLO instance related information, such as CLO ID, process instance ID, activity ID and
activity data. A constructor of the class is the only class method specified in the event
class.
A rule, which is also implemented as a Java class, consists of a condition
specification, an action specification, and alternative action specification. Each of the
condition, action and alternative action of the rule is implemented as a method of the rule
class. The data attributes of the rule class also include the CLO instance related
information as in the event class (i.e., CLO ID, process instance ID, activity ID and
activity data) so that the CLO instance information of an event occurrence can be passed
to an invoked rule. In addition, the rule class has an attribute, the value of which is a
reference to the LMSEngine. Through this reference, the action and alternative action
parts of a rule can invoke the methods of the LMSEngine.
A trigger is a Java object that links an event to a rule in order to allow the ETR
server to execute the rule when the associated event is posted. A trigger object specifies
the event that will activate the trigger, a rule that will be invoked by the trigger, and the
mapping between the parameters of the event and the parameters of the rule. The
parameter mapping information allows values of the data attributes of the event to be
transferred to the data attributes of the rule. During the CLO deployment process, the
code of the event and rule classes are generated, compiled and installed in the ETR
Server. A trigger is not code-generated, but a trigger object is instantiated with the
information specified above. It is registered with the ETR Server after instantiation.
When the Learning Process Execution Engine posts an event to the ETR Server during
the execution of a CLO instance, the ETR Server identifies the trigger object associated
with the event. Using the specification in the trigger object, the ETR Server identifies the
associated rule, passes the values of the data attributes of the event to the attributes of the
associated rule, and invokes the condition method of the rule class. The condition method
invokes either the action method or the alternative action method of the rule class
depending on the result of the condition evaluation. The action and alternative action
methods can identify the activity, the CLO instance, and the status of the activity by
referring to the data attributes of the rule class. These methods can in turn invoke the
methods of LMSEngine for adaptive, flexible and customizable delivery of e-learning
services.
CMC Tools
Interesting event occurred. (E-mail, discussion
Boards, etc)
Post Event(s) Schedules a c ference
l N e )( Te ISend e-mail
Notifies Executes Rule process Management
/ System
Call PIs for
oper ions
IT System
Components
Figure 5-11. Event and Rule Mechanisms for Coordination and Collaboration
In addition to sequential rule processing, the general ETR mechanism allows any
party interested in an event to subscribe to the event and receive a notification when an
occurrence of the event is posted. When an event is posted at a computing node, all
subscribers' nodes are notified, and triggers and rules associated with the event are
executed in a distributed fashion by the replicas of the ETR Server installed at the
subscribers' sites. We exploit the event notification mechanism to support collaboration
in e-leaming by integrating it with Computer Mediated Communication (CMC) tools. An
event is posted when a situation that requires collaboration has occurred. Rules triggered
by each instance of the event can perform collaborative operations such as sending e-
mails to collaborators, posting message to a discussion board, initiating a workflow
process, or arranging a conference call with fellow learners, the content author, and/or a
mentor. The event and rule mechanisms enable timely collaboration among people
involved in e-learning. Figure 5-11 shows the event notification mechanism that
facilitates collaborative e-leaming through the activation of CMC tools, workflow
processes, and other IT systems.
5.2.3.3 Assessment component
The Assessment Component performs the model-based assessment. It carries out
the assessment process, which involves selecting assessment items, collecting selected
assessment items from the activities participating in an assessment, grading the answers
submitted by learners, and recording the assessment result in the runtime status model of
the CLO instance that conducts the assessment.
An assessment in a non-leaf activity is carried out by making use of the assessment
items specified in the non-leaf activity as well as all or part of the assessment items of all
its child activities. The assessment component makes use of the Parsed CLO (internal
tree structure representation of a CLO) to identify the child activities of a non-leaf
activity. Additionally, assessment items in non-leaf activities can be retrieved from the
Parsed CLO.
A class named Assessmentlnfo is responsible for collecting and providing
assessment items, grading the answers provided by a learner and assigning measures of
the assessment to learning objectives. The information of each assessment item is
maintained in an object of the class named Assessmentltemlnfo. The collected assessment
items are stored in a vector named vAssessmentltemlnfo. The assessment items in the
vector are delivered in their indexed order. A hash table named htAssessmentltems is used
to store the vectors associated with learning objectives. The table uses learning objective
id as the key. Since assessment items of each objective are stored in a separate vector, the
measure of an objective can be easily calculated.
In case a leaf activity is bound to a CLO, an AssessmentInfo object of the CLO is
instantiated and stored in a vector in a hash table, htAssessmentlnfo, which uses learning
objective id as the hash key. The Assessmentlnfo object is also stored in the
vAssessmentlnfoOfBoundCLO vector for presentation. Assigning measures to objectives,
obtaining the bound LOs for collecting assessment items, and so on, are accomplished by
communicating with the LMSEngineCore object. When a learner submits an answer to an
assessment item, the user interface component would communicate with the RMI server
of LMSEngine (LMSEnginelmpl) to record the answer. The runtime structure of the
assessment component is illustrated in Figure 5-12.
1 1 1
1 1 1
Parsed CLO LMSEngineCore LMSEnginelmpl
S(RMI Server)
Figure 5-12. Runtime Structure of Assessment Component
5.2.3.4 Constraints-based LO broker
The Constraints-based LO Broker, which is an extension of UDDI v2.0 registry,
supports registration, browsing and dynamic binding of LOs. To register an LO with the
Broker, the RegBroker (QueryBroker) Web service operation should be invoked. The
operation requires the URL of an extended WSDL document describing its registration
(query) constraints as a parameter. The extended WSDL document includes the meta-data
description as well as the general WSDL elements, as discussed.
10. Obtain the LO Specification
(End of binding)
LPEE -
Business
service Keys
5. Discovery (Dynamic 8. Returns service
Binding) request with a keys, Access Point,
service constraints document ke, s of LOs
Constraints- 9. Request the
6. Search Web Services based Broker LO for Binding
(LOs) using business through Access
service keys Point with LO ID
7. Returns Business
Service structure
^ UDDI containing Access Points Web, SOAP Serer
Registry and IDs of the LOs (Acess Point
Sof LO (Only Once)
Extended WSDL (LO) Described LO
Service Interface By
Operations
Service Implementation 4. Register the LO with
2. Register service Access Point (URL) Meta-data & business key
implementation of Meta-data (Constraints) (End of registration)
an LO
J1: Business Service Key
3. Return a business service
key for the LO
Figure 5-13. Registration and Discovery Process of LO
LO authoring tools register LOs with the Broker, and take the following steps to
accomplish the task. First, an LO Web-Service interface is registered as a service
interface with the UDDI registry. The registration operation returns an interface key
(i b,,lel key). The interface can be considered as a business type that is common to all
LOs of that type. So, the interface registration is a one-time process. Then an LO, which
is an implementation of the service interface, is registered with the UDDI registry as a
business service, and a business service key is returned to the authoring tool when the
registration of the LO is completed. Together with the interface key and the business key,
the meta-data and constraints of the LO are registered with the Constraints-based Broker,
which serves as a meta-data repository having a constraint-matching facility.
The LPEE can discover an LO for dynamic binding using the constraint-matching
facility provided by the LO Broker. During the processing of a CLO instance, the LPEE
would issue a discovery request to the LO Broker by submitting an extended WSDL
document that includes a set of requirements. The constraint-matching facility would
match the specified requirements against the meta-data and constraints of LOs that have
been registered as Web services. The LO Broker finds a set of service keys of the LOs
that match with the requirements. Using the service keys, the LO Broker accesses the
UDDI registry to obtain LO access information, which includes the IDs and access points
of the LOs. The Web user interface retrieves the meta-data of each LO from its LO
repository to display the information to a learner to help him/her choose the right LO.
Finally, the LPEE accepts the learner's choice, and then invokes the getLO operation at
the access point of the chosen LO (i.e. the LO repository) with the ID of the selected LO
as an input parameter. The getLO operation fetches the LO and returns it to LPEE for
delivery. In Figure 5-13, steps 1 through 4 describe the registration procedure, and steps 5
through 10 describe the dynamic discovery and binding process. In case of a static
binding request, the LO Broker is given the business service key of a pre-selected LO
instead of service requirements in step 5. The LO Broker directly passes the business
service key to the UDDI registry without constraint-matching to obtain the service access
information. The WSDL document that describes the LO Broker Web service is given in
Appendix D.
5.2.3.5 User interface component
The User Interface Component is a set of Java Server Pages (JSPs), which generate
Web pages for communication between users and server components, such as LPEE,
Assessment Component, Monitoring Component and Constraints-based Broker. Through
the Web user interface, a learner can start, suspend, finish, navigate, and monitor a CLO
instance, and search LOs using meta-data constraints.
MainFrames JSP carries out user authentication and loading of a main Web page.
CLO instance commands, such as instantiation, completion, suspension, and resumption
of a CLO instance, are handled by cloCommands JSP. ActivityCommands JSP processes
the activity navigation commands, such as Continue, Previous Contents, Next Activity
and Previous Activity. Communication with the Assessment Component for assigning
answers to assessment items is accommodated through assessmentCommand JSP.
Monitoring JSP provides the status information of a CLO instance. Web user interface
that facilitates LO discovery through the Constraints-based LO Broker is provided by
searchLOs JSP. A user space called 'MyCLOs' can be accessed through a Web page
generated by myclos JSP.
5.2.3.6 Web user interface
A learner interacts with the Learning Process Management System through Web
pages that the User Interface Component generates. After a learner logs in the system,
79
he/she is able to search and execute LOs, access his/her learner space 'MyCLOs', or log
off.
, .- .
*j O .- I tv I
1.
Activit Command
C .n I E- I *iA I-, I
PIeCRe eeni hie consisoinl n frrrimoionDf lronaIng cije
youwrlh tO leaInm rT
for each corntroint aturLIeiDecIv The reIowIlr
infn moorn
.a-. ie '4,',. T.p Ve.-wd 4ou.-e'
I-- _1]1 I- II T
.w- _1F1- p ..
. --- _t1 -- 1p --
CLO C~mnd.
fiji- J
i -
~J RNvI I ~ I I ~I
Learning Objects in Progress or Finished
kSiftICt k* Jltowetrfflj
v M 1m iote74fdlS ,^o 01 enatu r- Ur)mI*
Learning Objects n MyCLOs
0I Vuori K.iy VInII. 1w
St'O,12B27174 .fl & l O I I t i> 1
1 CA0.1aw L37 A I Ifil c
Vsmncraa Itllr=
l wae ea '"n I
Cost eq 0.0
Idi-apla25a2i7Wuiaweau ] ]g "q uclle
f i. l ( I 1 141
%,row u?6
*W a-5 I C- _101
. j 1o1 21
, I
i,/
Legend
+ ONon-leafActivity DLeafActivity
*Color Code
+ 0 Current + 4 Suspended
+ Satisfied + *Unsatisfied
+ Unattempted + ` Disabled
- Current Status of CLO with Instance ID -
VAM1034bb5 103e74f3d01 -7ff0
= U VAM (Active)
Description
This is Root activity of Virtual
Anesthesia Machine instruction
Objective Status [id:Status(Norm
4Measure)]
-Overall Unknown (0)
Cmluent Status: Active
# of Attempts: 1
D Partly 1 asicConcepts (Satisfied)
M Part_2_IO_of VAMI (Attempted)
+ Part 3_SafetyExercises (Current)
] .: :. C DE | Interet l
Figure 5-14. Web User Interface. A) Search Constraint Specification. B) 'MyCLOs'
Learner Space. C) CLO Status Monitoring Information.
trC I' 9l-. 1U yc
LQ2-urv IcP
CLO Commands
SI 0.0
F." CLO
,lll.r~.,,..~m.m.r~r~u* -~-lilP IIII
c--m,,~;mnimruc EY~
13 urr 5atmof L.B- -loor I
- F- r
r 1,*.-..
]
When a learner chooses to search LOs, a page that accepts search requirements is
provided as shown in Figure 5-14 (A). A list of LOs that meets the requirements and their
meta-data is displayed as the result of the search request. In the search result page, LOs
can be added to 'MyCLOs' for later execution, or an LO can be instantiated for execution.
In 'MyCLOs' Web page, a learner is able to start an LO registered in 'MyCLOs', gain
learning experience by executing the LO, suspend, finish, or resume existing LO
instances. Figure 5-14 (B) shows the 'MyCLOs' page. During the execution of a CLO
instance, a learner can monitor the status of a CLO instance through a Web browser as
shown in Figure 5-14 (C).
CHAPTER 6
APPLYING LEARNING OBJECT AND DYNAMIC E-LEARNING TECHNOLOGIES
TO SIMULATION-BASED MEDICAL INSTRUCTION
There are many areas of education where our learning object models and e-learning
service infrastructure can be applied. As a part of this work, we aim to apply the e-
learning technologies to a practical area where there can be a direct impact.
6.1 Instruction on Anesthesia Machines
Complex medical equipment is regularly used in hospitals and clinics where patient
safety depends on the proper interaction between skilled practitioners and equipment
[Dalley, et. al., 2004] [Reason, 2000]. Even though anesthesia is increasingly safe,
critical accidents in which patients are seriously injured still occur [Dalley, et. al., 2004]
[Weinger, 1999]. Misuse of equipment is far more common than pure equipment
errors/failures in the medical environment. Human error is a dominating factor in up to
90% of the problems caused by equipment [Weinger, 1999] [Williamson, et. al., 1993].
There are several Web contents that provide learning materials on anesthesia
machines. For example, the resource reported in [Dosch, 2004] provides extensive Web
contents about a variety of anesthesia machines. The material by the College of
Veterinary Medicine [Washington State University, 2004] is a comprehensive resource
that covers circuits, vaporizers, gas cylinders, pressure regulators, flow meters,
scavenging, ventilators, and endotracheal tubes. However, they do not provide simulation
software for practice, nor a way of assessment to measure learners' understanding of the
materials. [Gas Man, 2004] is a commercial product for teaching, simulating and
experimenting with anesthesia uptake and distribution. It provides a text tutorial and
simulation software. The text tutorial is not accessible on the Web; it is published as a
book. The simulation software is not implemented as a learning object. To the authors'
knowledge, no anesthesia simulation system has been developed based on the learning
object technology at the time of writing.
6.2 Introduction to the Virtual Anesthesia Machine
There are several ways for practitioners and medical students to learn how to use an
anesthesia machine. First, a person who is proficient in the use of the machine can
provide a training session by giving a demonstration and using textbooks and reference
materials supplied by the manufacturer of the machine. Second, a practitioner can learn
by using the anesthesia machine with a Human Patient Simulator (HPS) [Dalley, et. al.,
2004]. Third, a Virtual Anesthesia Machine (VAM) [Lampotang, et. al.] that simulates a
real counterpart can be used for educational purposes. VAM is a Web-based simulation
system developed at the University of Florida. It, together with a Web-accessible
workbook, can be accessed by registered users free of charge through Web-enabled
personal computers. VAM is an interactive simulation of a generic anesthesia machine
with model-driven, real-time animation of color-coded molecules and dynamic
representations of flows, pressures, volumes and concentrations of gases. It is
international in scope and features legends in 23 languages and is used in over 336
institutions and programs worldwide. The VAM web site at http://vam.anest.ufl.edu
receives more than 2,000,000 hits per year.
A major advantage of VAM is that users can focus on learning about anesthesia
without being distracted by the concern of placing an actual patient at risk [White, 2002].
Also, it is a cost-efficient way to learn because it is free to use and does not require any
actual anesthesia equipment or patient. In addition, practitioners and medical students can
learn the inner workings of the anesthesia machine because VAM is a transparent reality
simulator; that is, it is a model-driven, display-based simulation that represents internal,
abstract and invisible functions with explicitly visible and manipulatable symbols to
assist learners in exploring, developing and confirming mental models. The focus of the
simulation is on high fidelity behavior rather than life-like appearance [Lampotang, et.
al.].
In spite of VAM's success and advantages, there are a number of limitations in the
existing system. First, it is a monolithic system in that the components and their
accompanying instructional materials, questions and answers for practice and questions
for assessment are not modularized to make them reusable for constructing other
simulation systems. Second, users of VAM and its workbook can freely operate on any
part of the system and access any content of the workbook. Although, the workbook
provides some guide for structured exercises, the learners can choose not to follow the
guide. Thus, the last of the four learning approaches (receptive, directive, guided
discovery and explorative) proposed by Dr. Ruth Clark [Clark, 2000] is emphasized.
Some learners may desire or need more structured learning approaches than the
exploratory approach. Lastly, the ability of the existing VAM system to adapt and
customize its instruction delivery to suit the learners' profiles and needs is currently
limited to language and medical gas color code. Searching for contents on the VAM Web
site is accomplished via a search box, driven by Google, which is a part of the header of
every Web page on the VAM Web site. This chapter discusses the application of the
learning object and e-learning service technologies we have developed to address these
limitations.
6.3 Application to Medical Instruction using VAM
The Virtual Anesthesia Machine (VAM) simulates the inner workings of an
anesthesia machine and ventilator the complex equipment in which oxygen, nitrous
oxide and anesthetics mix together to provide patients insensitivity to pain during surgery
or procedures. The objective of VAM is to increase the safety of patients by providing
practitioners clear understanding of the flow of gas in the machine, user actions and the
consequences of equipment malfunctions [White, 2002].
VAM is composed of a transparent reality simulator developed using Macromedia
Director, and a Web-based workbook (http://vam.anest.ufl.edu/members/workbook/apsf-
workbook-english.html) delivered in the Macromedia FlashPaper format. The simulator
currently supports 22 languages, and the workbook is in 11 different languages. In the
visual display of the simulator, the language of the user interface as well as the color code
for medical gases are changed according to a learner's selection or data retrieved from a
user database.
The workbook consists of three parts. Part 1 covers the basic concepts of the
anesthesia machine. Part 2 teaches input/output controls of VAM. Users can adjust gas
pipelines, valves, gas flow meter knobs, buttons, and so on, and visually observe the
effects of the adjustment on the flow of molecules of gases, the readings of pressure
gauges, valves, and so forth. Part 3 covers safety exercises and is composed of six
sections for covering the high pressure system, the low pressure system, the breathing
circuit, manual ventilation, mechanical ventilation, and the scavenging system,
respectively. Each section is comprised of a set of learning objectives, a set of
instructional materials, and a set of post-assessment questions. Each instructional material
has a question, an answer, a VAM demonstration instruction, an explanation of the
demonstration, and some learning objectives. The post-assessment questions (quiz) can
be answered online at a Website specified in the workbook.
The workbook, the simulator and the quiz Web site need to be used together for
effective delivery of learning experience. However, the current VAM system is not
structured to provide a smooth transition between the workbook contents, the simulator,
and the post-assessment questions. Even though directions are provided in the workbook,
it is the responsibility of a learner to open the workbook online, to start the simulator, to
go to the quiz Website at appropriate times. The learner is free to do anything with the
contents and the simulator (i.e., explorative learning). The learning process is generally
not guided by the system.
6.3.1 Encapsulating VAM as Learning Objects
There are two main advantages to create LOs that encapsulate the workbook
contents, the VAM simulator, and the quiz problems. First, each LO can be reused for
instruction of other types of anesthesia machines, which may use the same or similar
componentss. Second, it is possible to deliver learning experiences provided by the
workbook's contents, the use of the simulator for practice, and the use of quiz problems
for assessment in a seamless fashion and under the control of the Learning Process
Management System.
In this application scenario, we created an ALO for Part 1 and another ALO for
Part 2 of the workbook. Part 3 is composed of several ALOs each of which contains
instructional materials to cover a component of the anesthesia machine. For these ALOs,
the workbook contents are extracted and regenerated as separate HTML pages. The quiz
86
problems available at the quiz Website are extracted and used as post-assessment items.
We constructed a CLO for modeling the entire VAM instruction as an activity tree, in
which each of the leaf activities is bound to an ALO. Figure 6-1 shows the activity tree of
the CLO, and the relationship among the ALOs and VAM resources.
VAM Instruction CLO
r---------------------
S. ...41Rds to
-ORL
,l w Assets Created
Sa by Referencing Materials
............. A r" r'
.... .. ..... ....,. **."' .
S .......... l... ............ B
Content items
HTML
Pages
Workbook Simulator Quiz Web Site VAM
A ... Learning
S.. 1 Resources
Figure 6-1. Construction of LOs using VAM Learning Resources
An ALO bound to a leaf activity of Part 3 begins its presentation with a textual
description of learning objectives of the ALO. Then a sequence of several content and
practice items of the workbook is delivered. A content item is presented to the learner in
the form of Q&A as in the VAM workbook. For example, a content item in the High
Pressure System ALO is as follows: "Question: During the morning pre-use check of an
anesthesia machine, you open the 02 cylinder to check the 02 cylinder pressure. The 02
cylinder reads full (2,000 psig). You subsequently close the 02 cylinder. With the anesthesia
machine turned on with a minimal flow of200mL/min of 02, what does the 02 cylinder pressure
gauge now read with the 02 pipeline supply connected and the 02 cylinder closed? Answer:
87
Even ;, li,,,li the 02 cylinder has been closed, the 02 cylinder pressure gauge will continue to
read full."
I .-2. 1 .
It",
Welcome glee
M CLs
Search CLOs
CLO Commands
Suspend CLO
Fnish CLO
SI I
3 V' ~ a 1
-------- ~~.i2a !. --
Activity Commands
Prvotn. J t Ex.tAc. t.I P1riousAcdtly I NIA.t.v. I
Practice Item to be presented.
Answer is on the bottom of the page.
Please click the link below to see the learning item.
I I I
I_ J Ir I-1 ": __ ___o ''l I ; :
I 21.1 I- I I
lJ
Demonstration using VAM s ru aton (coordinates reer to VAM grand map below)
Pes Reet (E13 to start imulaton aresh without caLyoerro preiou ntentions
a Chcl on vaporizer concentration dia con () to swch vaporizer to a nonzero setting The external wofthe
vPepoz erfades mio a cul-ou. vew when the cutsor rs placed on ihe vap orzer Ico FGF moleculs fbow over the
hqud anesthet c pool insde he vapozer and pc up vaporized anesthet c molecules
b Press the 0, flush icon (D-B) The fiow 0O, molecules caused by pressing the 0C fish flows dr-eciy ino he
breathing c.rcut without p-assmg through the vaporzer The molecules from the C2 and N2O fjwmetecs contmue to
flow into the breathing circuit during an 02 flush
Explanation The high 02 flow (approx 60 Umln) generated by an 0, flush does not flow through the vaporizer and is thus at a
volatile anesthetic concentration of 0% The 0, flush thus dilutes the anesthetic concentration in the breathing circuit
iij,,L--
k r J 1V-1
Figure 6-2. Web User Interface of VAM Instruction. A) E-Learning Service System
Delivering Practice Item. B) A Practice Item with Demonstration Scenario.
