The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
EDUCON 2010: Adaptation in a PoEML-based E-learning Platform
1. Adaptation in a PoEML-based E-
learning Platform
Roberto Perez-Rodriguez
Manuel Caeiro-Rodriguez
Luis Anido-Rifon
2. Introduction
• Educational Modeling Languages (EMLs) enable
the definition of UoLs in a pedagogy-independent
way.
• An EML describes not only contents, but also
– Activities to perform
– Composition of groups of participants
– Order between activities
• These specifications conform a process.
• The PoEML language follows into this conceptual
model, and it was developed following a
separation-of-concerns approach.
3. Life-cycle of a Unit-of-Learning
• Authoring: in this phase,
the UoL is designed by
making use of an
authoring tool.
• Publication: the UoL is
imported into an
execution engine.
• Delivering: the
participants perform the
UoL. This phase is also
known as run-time.
6. Late Modeling
• A main issue in the design of e-learning systems is that
of the adaptation.
• Late modeling allows for designers to leave some parts
of the UoL unmodeled until run-time.
– Often, designers may prefer to model the most relevant
parts of the UoL, remaining the rest to be modeled at run-
time.
• Therefore, it is required to support incomplete (or
abstract) models, which will be specified in run-time.
• This paper introduces a conceptual framework and a
software architecture for supporting late modeling of
PoEML learning-flows.
7. Common Approaches to Adaptation
• Approach centered in process definition meta-
model
– Advanced modeling
– Late modeling
• Approach centered in the execution system
– Type adaptation (evolutionary change)
– Instance adaptation (ad-hoc change)
8. Approach Centered in Process-
Definition Meta-Model
• In this approach, it is the process meta-model the one that
determines the structure and the allowed types of changes.
The following lines are considered:
– Advanced modeling
– Late modeling
• Advanced modeling allows authors to consider alternatives
during run-time.
– The different alternatives (learning paths) that are allowed
during execution are modeled in the process definition at
design-time.
– During run-time, it should be possible to select one of those
alternatives. There are two main ways to do the selection:
• In accordance with the state of execution.
• In accordance with the decision of an authorized participant.
9. Approach Centered in the Process-
Definition Meta-Model (ii)
• Late modeling allows for leaving some parts of
the design without being modeled, in order to
be modeled during run-time.
• Some parts of the process definition are
leaved as black-boxes, in order to be modeled
during execution.
10. Approach Centered in the Execution
System
• In this approach, it is the execution system the
one that supports the types of change that are
allowed.
• The following lines are considered:
– Instance adaptation
– Type adaptation
11. Approach Centered in the Execution
System (ii)
• Instance adaptation is usually named exception
management. An exception can be seen as an
occasional deviation from the normal behavior of
a process. Exception management is in charge of
correcting the wrong situation, without need for
stopping all process instances, fix the model, and
enact all the instances again.
• Type adaptation deals with process instance
migration during a process execution from an old
schema to a new one.
13. A Late Modeling Example (ii)
• The initial scenario.
• Once the initial
scenario is finished,
three scenarios are
launched in parallel.
• Finally, participants are
requested to perform a
collaborative activity,
which is evaluated by
the teacher.
14. A Late Modeling Example (iii)
• Order and Temporal
views of scenarios
• The teacher may want
to tune the model in
some ways:
– Adding more activities.
– Changing the temporal
constraints for activity
finishing.
– Adding/skipping a
goal, etc.
15. Late Modeling of PoEML Learning-
Flows
• Following an aspect-oriented approach, the
allowed changes in the model can be
systematically categorized in aspects, each
one dealing with an specific concern.
• A. Situating placeholders in the UoL model
– The approach here is to use a special type of
construct, named placeholder.
– The placeholder occupies the place of the
unmodeled element or specification.
16. Late Modeling of PoEML Learning-
Flows
– A. Situating placeholders in the UoL model.
– In PoEML, we identify the following placeholder types:
• Scenario placeholder. It is a black box that reserves space for a
scenario model. The language constructs will be added at run-time
in the space kept by the placeholder.
• Goal placeholder. It reserves the space for the definition of a
learning objective.
• Environment placeholder
• Tool placeholder
• Data placeholder. It is used for reserving the space to add
resources and/or activities to a UoL at run-time.
• Order placeholder. It keeps the space for adding ordering
constraints at run-time.
• Temporal placeholder. It enables the deferred definition of
temporal constraints in the Unit of Learning.
17. Late Modeling of PoEML Learning-
Flows
• B. Modeling black-
boxes at run-time
• When the execution
flow reaches a black-
box, the educational
scenario is
instantiated, but it
cannot progress
towards the accessible
state.
18. Late Modeling of PoEML Learning-
Flows
• A learning-flow with a
temporal placeholder
in the final scenario.
• At run-time the
teacher may fill the
placeholder by making
use of the authoring
environment.
19. Late Modeling of PoEML Learning-
Flows
• C. Updating of instances.
• Every learner who starts a learning-flow
enacts a new instance.
• Every learning-flow instance has its own state.
• When a teacher commits a change in the
learning-flow model, every instance has to be
re-evaluated from root to leaves in order to
update its state.
20. The Moodle Extension for PoEML
• A. The authoring environment
– The authoring process consists on a series of
atomic operations.
• Add/edit/delete a scenario
• Add/edit/delete a goal
• Add/edit/delete a resource/activity
– This authoring strategy presents some advantages
• Atomic changes are automatically seen by other co-
authors
• There is no need for a complex consistency check like
the one needed when importing a manifest file
22. The Moodle Extension for PoEML
• C. Monitoring
– Monitoring of a goal
• Possible states of a goal and number of students per state.
Histogram
• Detail of the status of a goal for every enrolled student
24. Implementation
• A. The learning-flow engine
– It is the core component of the e-learning system
– Stores information on scenarios, participants,
goals, etc. and it makes the state of the system to
evolute depending on events
– This component is integrated through a well-
defined interface
– Two subcomponents:
• Models manager
• Instances manager
25. Implementation
• B. The middleware layer
– The functionality of the learning-flow engine is
published into a WSDL file.
• C. Presentation modules
– Three views:
• Authoring
• Monitoring
• Delivering
26. Conclusions
• This paper details the implementation of a late
modeling approach for PoEML
• Late modeling allows for leaving some parts of a UoL
model unspecified until run-time
• The aspect-orientation of PoEML presents great
advantages in supporting late modeling
• The combination of late modeling and advanced
modeling provides a high degree of adaptation
• Future work: to support adaptation based on the
execution engine
– Instance adaptation
– Type adaptation