SUPPORTING LEARNER MODEL EXCHANGE

IN EDUCATIONAL WEB SYSTEMS

Eddie Walsh, Alexander O’Connor and Vincent Wade

Knowledge and Data Engineering Group, School of Computer Science and Statistics

Trinity College Dublin, Dublin 2, Ireland

Keywords: Interoperability, Mapping, Educational web systems, Learner models, Adaptivity.

Abstract: The heterogeneity of learner models in structure, syntax and semantics makes sharing them a significant

challenge for existing educational web systems. Creating mappings between the different types of learner

models is one technique that is used when attempting to overcome these issues. This paper presents an

overview of research currently being conducted in the area of learner model exchange and defines a

categorization, derived from existing educational web systems, of the different mapping types that are

required for learner model mapping. Following this, a framework is presented that supports the creation and

validation of these different mapping types and the exchange of learner information between multiple

heterogeneous educational web systems.

1 INTRODUCTION

Providing advanced levels of learner model

interoperability between different educational web

systems, learning management systems, learner

databases and educational administration systems is

a very difficult challenge due to high levels of data

heterogeneity at the structural, syntactic and

semantic levels (Cena & Furnari 2008). The

emergence of adaptive educational web systems that

often require rich forms of learner information to

support their personalization functionality further

complicates these issues.

Many different approaches have attempted to

provide learner model interoperability. However,

achieving a high level of interoperability between

learner models can result in very complex

integrations which currently cannot be fully

automated (Falconer, Noy & Storey 2007). One

method that is commonly used in data integration is

the creation of mappings. Mappings are associations

between equivalent data from different data model

representations. In educational web systems, the

mapping approach can be used to perform

translations of heterogeneous learner data between

independent educational web systems.

This paper presents an overview of the current

research in the field of learner exchange and defines

a categorization of different mapping types that

facilitate sharing between existing learner models.

These mappings are derived from the analysis of

learner information in a variety of educational web

systems including the two main open source learning

management systems; Sakai (Sakai 2010) and

Moodle (Moodle 2010), and a number of adaptive

educational web systems such as AHA! (De Bra et

al. 1998), APeLS (Conlan & Wade 2004),

CUMULATE (Brusilovsky, Sosnovsky &

Shcherbinina 2005).

Subsequently, a framework for the creation and

validation of these mappings and the automated, on-

demand exchange of learner information between

heterogeneous educational web systems is presented.

This framework, called FUMES, incorporates web-

based, domain-specific tools designed to aid the

administrator in carrying out the necessary

interoperability tasks, in particular, the creation of

mappings between heterogeneous learner data.

A case study to test the validity of FUMES in a

practical learner exchange setting has also been

performed. This scenario demonstrates the

application of FUMES in the domain of database

and SQL education.

346

Walsh E., O’Connor A. and Wade V..

SUPPORTING LEARNER MODEL EXCHANGE IN EDUCATIONAL WEB SYSTEMS .

DOI: 10.5220/0003348403460351

In Proceedings of the 7th International Conference on Web Information Systems and Technologies (WEBIST-2011), pages 346-351

ISBN: 978-989-8425-51-5

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

2 RELATED RESEARCH

There has been considerable effort put towards the

development of a standardized learner model, for

example IMS LIP (IMS 2010). However,

standardized learner models have failed to gain

widespread adoption due to the diversity of

educational web systems and their resulting learner

representations. This has led to significant research

into the field of learner model interoperability.

A variety of approaches to learner model

interoperability have been attempted. The most

common current implementations are hybrid

approaches that take aspects of both the centralized

and distributed methods of learner model

interoperability (Van Der Sluijs & Houben 2006)

(Bielikova & Kuruc 2005).

At the syntactic level, the use of a standard

language representation such as XML (De Bra et al.

1998) (Conlan & Wade 2004) is the most common

with some approaches now adopting semantic web

technologies such as RDF or OWL (Van Der Sluijs

& Houben 2006) (Bielikova & Kuruc 2005) (Dolog

& Schäfer 2005). A standard transfer protocol, such

as web services, is also widely used (Bielikova &

Kuruc 2005) (Cena & Furnari 2008).

At the semantic level, some approaches have

attempted to adopt compliance with a common

canonical learner model or learner model server

across all integrated systems (Heckmann et al. 2005)

(Dolog & Schäfer 2005). Other approaches have

attempted to reconcile heterogeneous learner models

by using mapping or mediation techniques (Van Der

Sluijs & Houben 2006) (Bielikova & Kuruc 2005).

The mapping approach, which has been used

extensively for interoperability in many other related

fields such as databases (Sheth & Larson 1990) and

more recently ontologies (Kalfoglou & Schorlemmer

2003), provides significant potential for learner

model interoperability. However, in many cases

mappings are provided manually without tool

support or by employing one of the current generic

schema or ontology mapping tools available such as

COMA++ (Aumueller et al. 2005). These mapping

tools cannot provide a systematic approach to

learner model mapping and many have been found

to be too general, built without domain-specific

mechanisms, lacking in visual displays or easy to

use components and not allowing for expressive

enough mappings (Falconer, Noy & Storey 2007).

To execute an exchange of learner data, some

approaches offer a pre-runtime, administrator-

initialized process (Dolog & Schäfer 2005) while

others can perform the exchange in a runtime, on-

demand process (Heckmann et al. 2005) (Van Der

Sluijs & Houben 2006) (Vassileva, McCalla &

Greer 2003). Most approaches provide support for

multiple learner models interoperability scenarios.

However, only a few of the approaches reviewed

provided details on how they reconcile conflicting

and incomplete learner data (Van Der Sluijs &

Houben 2006) (Bielikova & Kuruc 2005).

3 LEARNER MAPPING

CATEGORIZATION

From examining sample learner information from a

variety of educational web systems such as Sakai

(Sakai 2010), Moodle (Moodle 2010), AHA! (De

Bra et al. 1998), CUMULATE (Brusilovsky,

Sosnovsky & Shcherbinina 2005) and APeLS

(Conlan & Wade 2004) a set of required learner

mapping types have been derived and categorized.

These learner mappings consist of core generic

mapping types that are combined as needed to allow

the exchange of heterogeneous learner information.

In the following sections, these mapping types are

explained and examples of each, in the domain of

database and SQL education, are given in table 1.

3.1 Core Generic Mapping Types

Equivalence schema mappings are the most basic

form of mapping and are created between the

equivalent schema elements of learner models. Two

extensions of this type of mapping are the join

schema mapping, where multiple schema elements

from one learner model are equivalent to one

schema element in another learner model, and

separation schema mapping, where one schema

element is equivalent to multiple schema elements.

Building on the schema mapping, functional

mappings allow generic manipulation of instance

data in an exchange between learner model schema

elements. Types include numeric mappings which

allow mathematical manipulation of numerical data,

format conversions which allow manipulation of

data types such as dates, and interval mappings

which allow manipulation of data that requires the

use of intervals, for example, learner grades.

Next, equivalence instance mappings allow more

complex mapping of specific instance data from

learner model schema elements. An example would

be the matching of equivalent user IDs that are

represented differently in heterogeneous systems.

SUPPORTING LEARNER MODEL EXCHANGE IN EDUCATIONAL WEB SYSTEMS

347

Table 1: Examples of Mapping Types.

Mapping Type Example Mapping

Equivalence Schema Autho

= Creator

Join Schema FirstName & LastNam

= FullName

Separation Schema Addres

= Street & City

Numeric Score[0.8]*10

= Result[80]

Format Conversion Date[2010-03-29] = Date[29/03/2010]

Interval Grade[A] = Percentage[90-100%]

Equivalence Instance UserID[jsmith] = UserID[06125]

Equivalence Domain Instance Concept[SQL_A] = Concept[SQL1]

Join Domain Instance Concept[SQL_A] & Concept[SQL_B] = Concept[SQL1]

Separation Domain Instance Concept[SQL_A] = Concept[SQL1] & Concept[SQL2]

Competency Concept[SQL_A] & (Score[0.8]*100

)

Concept[SQL1] & Progress[80]

Separation Competency Concept[SQL_A] & (Score[0.8]*100

)

(Concept[SQL1] & Progress[80]) &

(Concept[SQL2] & Progress[80])

Cross-category Assessment[SQL_Quiz1] & (Score[8]*10

)

= Concept[SQL1] & Progress[80]

Separation Cross-category Assessment[SQL_Quiz1] & (Score[8]*10

)

(Concept[SQL1] & Progress[80]) &

(Concept[SQL2] & Progress[80])

Instance mappings can also be created using a

predefined set of possible instance values, for

example, domain concepts retrieved from domain

models. Again, extensions of the instance mapping

type can include joining and separating.

3.2 Composite Learner Mapping Types

Individually, the generic mapping types are not

sufficient to map complex learner model data.

However, they can be combined to create mappings

between different categories of learner information.

The categories of learner information used in this

analysis are based on the IMS LIP specification

(IMS 2010). They are (i) Identification, (ii) Goals,

(iii) Interests, (iv) Assessments, (v) Competencies,

(vi) Activities, (vii) Qualifications, (viii)

Affiliations, (ix) Accessibility, (x) Security.

An example of a mapping in the competencies

category would be the instance mapping of a domain

concept in conjunction with the numeric mapping of

the learner’s knowledge of that concept. Join and

separation extensions also apply at this higher level

of learner mapping types, for example, exchanging

one competency from a learner model into two

competencies in another learner model.

Mappings can also be created between the

different categories of learner information. These

cross-category mappings allow for very expressive

exchange of learner data. An example is equating an

assessment in one learner model to a competency in

another learner model. In cross-category learner

mappings, join and separation extensions are also

possible. For example, separation of an assessment

in one learner model into two competencies in

another learner model.

In summary, these learner mappings are a core

set identified using an evidence-based approach

where existing learner representations were analyzed

for potentially shareable information. The selected

systems are a representative sample of typical

educational web systems and use many common

user modeling techniques, such as the overlay

approach in adaptive systems (Brusilovsky & Millán

2007), that would likely be present in other systems’

learner models. However, there are potentially other

mappings to be found in learner data that is more

complex and less suitable for sharing such as in

event-driven user data or where there is

interdependency of data within a learner model.

4 FEDERATED USER MODEL

EXCHANGE SERVICE

Providing a means to validate the identified learner

mappings has led to the development of an

interoperability system called FUMES. FUMES

allows the creation of the various learner mappings

types and acts as a mapping execution environment

for the automatic exchange of learner information

between educational web systems. The FUMES

approach to interoperability consists of a pre-

runtime, administrator-led mapping process and a

runtime, automatic exchange process.

WEBIST 2011 - 7th International Conference on Web Information Systems and Technologies

348

Figure 1: FUMES Architecture.

The mapping process has led to the development

of the Learner Mapping Web Application; a

graphical tool for the manual creation of mappings

between learner models. After receiving sufficient

training, the administrator can used this tool to

perform a number of tasks as shown in figure 1.

(1a) The administrator uses the Learner Mapping

Web Application to import learner model schemas

from educational web systems.

(2a) The administrator uses the Learner Mapping

Web Application to create mappings between each

learner model and the FUMES canonical model,

which is based on IMS LIP (IMS 2010). Equivalent

selections are made from each model and the

mapping type is specified.

(3a) The Learner Mapping Web Application

automatically generates XQuery versions of these

graphical mappings which can be tested using

sample learner model instance data. When the

mappings have been verified they are stored in the

FUMES database for use at runtime.

When the administrator-led mapping process has

been completed for a number of systems, FUMES

can perform an automatic exchange of learner

information between those systems. FUMES uses

web services to allow access to heterogeneous

learner models and provides a means to transfer

them between different systems using a common

standard protocol. Currently, FUMES supports

learner models represented in XML; the most

commonly used format in existing educational web

systems. In the future, it may be extended to include

semantic web technologies such as RDF and OWL.

Within the FUMES framework, the central point

for exchange is the Learner Translation Web

Service. This service handles the management of the

learner model interchange and translates between the

various learner model representations. Again, figure

1 shows the stages in the exchange process.

(1b) The end-users access one of potentially many

educational web systems that retain learner

information about them.

(2b) Each system can request updates of its learner

model instances from the Learner Translation Web

Service. The Learner Translation Web Service

retrieves suitable learner model instances from other

educational web systems.

(3b) The retrieved learner models are each translated

and merged into the FUMES canonical model form

by executing the XQuery mappings in the database.

Finally, this canonical learner model representing

the various source learner models is translated into

the appropriate learner model form and returned to

the target educational web system.

5 CASE STUDY

A case study has been conducted using FUMES to

demonstrate the exchange of learner information in

the practical learning setting of database and SQL

education. The case study incorporates the two main

open source learning management systems; Sakai

and Moodle, the adaptive educational web system

APeLS and the learner modeling system

CUMULATE.

SUPPORTING LEARNER MODEL EXCHANGE IN EDUCATIONAL WEB SYSTEMS

349

Table 2: Case Study Results.

Source Target Mapping Types Mappings

Moodle (Learning

Styles)

FUMES

(Accessibility)

Accessibility Learner Mapping

(Equivalence Instance & Equivalence Numeric)

Sakai

(Assessments)

FUMES

(Assessments)

Assessment Learner Mapping

(Equivalence Instance & Equivalence Numeric)

CUMULATE

(Competencies)

FUMES (Cognition)

Cognition Learner Mapping

(Equivalence, Join, Separation Domain Instance & Numeric)

FUMES

(Accessibility)

APeLS

(Learning Styles)

Accessibility Learner Mapping

(Equivalence Instance & Equivalence Numeric)

FUMES

(Assessments)

APeLS

(Competencies)

Cross-category Learner Mapping

(Separation Instance & Separation Numeric)

FUMES

(Cognition)

APeLS

(Competencies)

Cognition Learner Mapping

(Equivalence, Join, Separation Domain Instance & Numeric)

Total Number of Mappings

Total Mapping Execution

Time (ms)

APeLS Initialization Time

(No Integration) (ms)

APeLS Initialization Time

(FUMES Integration) (ms)

36 581 502 1669

The learning management systems were chosen as

they are often the central point for online learning

and are extensively used by many institutions. The

adaptive system and learner modeling system were

chosen as they retain more complex learner

information to provide personalization to learners.

The case study consisted of sample learner

information from Sakai, Moodle and CUMULATE

being supplied to an APeLS-based adaptive web

course used to teach SQL. The adaptivity within the

SQL web course is supported by a learner model

generated from prior knowledge and learning style

questionnaires that new students complete before

using the system. The goal of this case study was to

identify if FUMES could support the necessary

mappings to allow the alternative retrieval of prior

knowledge and learning styles from Sakai, Moodle

and CUMULATE learner representations.

5.1 Implementation

To achieve the integration, the FUMES Learner

Mapping Web Application was used to identify

suitable mappings from the source learner

representations to the FUMES canonical model.

Mappings were then identified from the FUMES

canonical model to the APeLS learner model. The

mappings allowed the exchange of competencies

from CUMULATE, assessments from Sakai and

learning style information from Moodle. These

mappings were then stored for execution at runtime.

To execute the exchange of learner data, the SQL

web course was set up to request new learner models

from the FUMES Learner Translation Web Service.

When a request was received by FUMES, the

mappings were executed and a new learner model

was generated based on the competencies found in

CUMULATE, the assessments found Sakai and the

learning style information found in Moodle. This

learner model was translated into an APeLS learner

model representation and returned to initialize an

adaptive learning session for the learner.

5.2 Performance

An analysis of this case study was carried out to

examine the types of mappings required and the

performance of those mappings when executed to

exchange learner information.

Table 2 shows the results of mapping the

individual source learner models to the FUMES

canonical model and the FUMES canonical model to

the target learner model in the APeLS-based SQL

web course. Table 2 also shows the overall

performance results for the full integration of

Moodle, Sakai, CUMULATE and APeLS using

FUMES. The total number of mappings created in

this case study using the canonical model approach

was 36 and the total execution time for all the

mappings was 581ms. The key result of this case

study is the time taken by the APeLS-based SQL

course to receive its updated learner model from the

FUMES Translation Service and instantiate a new

learning session for the learner. If this task is slow

there will be a negative impact on the usability of

the SQL web course.

Table 2 shows that the use of FUMES to retrieve

updated learner models adds just over one second to

WEBIST 2011 - 7th International Conference on Web Information Systems and Technologies

350

the initialization time of the APeLS-based SQL web

course. This should not impact greatly on the

usability of the system. Another major benefit of

using FUMES is the removal of the prior knowledge

questionnaires for first-time learners if that prior

knowledge can be retrieved from other systems. This

could potentially save significant time previously

spent completing extensive questionnaires.

In summary, this was an initial case study to test

FUMES in a practical learner exchange scenario. It

demonstrates that FUMES can successfully support

multiple integrations between heterogeneous

educational web systems. It has also shown that the

shared learner information can be used successfully

by the target system, in this case, to automatically

personalize content for new learners. The

performance results indicate that, in this case study,

the mapping approach is viable and does not

significantly decrease the responsiveness and

usability of the integrated systems. However, further

research will be required into the scalability of this

approach for larger numbers of mappings.

6 SUMMARY

This paper has given an overview of current research

in the area of learner model interoperability and has

defined an evidence-based categorization, based on

existing learner models, of mapping types that are

required to perform learner model mapping.

To validate these mappings, an interoperability

system was implemented to support the sharing of

heterogeneous learner information. FUMES allows

the creation of complex relationships between

multiple learner models, using a visual approach to

resolve domain-specific mapping problems.

Finally, a case study demonstrated FUMES

supporting the mapping and exchange of learner

information between multiple existing educational

web systems with minimal impact on usability.

REFERENCES

Aumueller, D., Do, H-H, Massmann, S & Rahm, E 2005,

Schema and Ontology Matching with COMA++, In

2005 ACM SIGMOD International Conference on

Management of Data. ACM.

Bielikova, M. & Kuruc, J., 2005, Sharing User Models for

Adaptive Hypermedia Applications, In ISDA 05, 5th

International Conference on Intelligent Systems

Design and Applications. IEEE Computer Society.

Brusilovsky, P. & Millán, E., 2007, 'User Models for

Adaptive Hypermedia and Adaptive Educational

Systems', in The Adaptive Web, Springer, pp. 3-53.

Brusilovsky, P., Sosnovsky, S & Shcherbinina, O 2005,

User Modeling in a Distributed E-Learning

Architecture, In User Modeling 2005. Springer.

Cena, F. & Furnari, R., 2008, A SOA-Based Framework

to Support User Model Interoperability, in 5

International Conference on Adaptive Hypermedia

and Adaptive Web-based Systems. Springer.

Conlan, O. & Wade, V., 2004, Evaluation of APeLS-An

Adaptive eLearning Service based on the Multi-model,

Metadata-driven Approach, In 3

International

Conference on Adaptive Hypermedia and Adaptive

Web-Based Systems. Springer.

De Bra, P., Aerts, A., Berden, B., de Lange, B., Rousseau,

B., Santic, T., Smits, D. & Stash, N., 1998, AHA! The

Adaptive Hypermedia Architecture, The New Review

of Hypermedia and Multimedia. vol. 4, pp. 115-139.

Dolog, P. & Schäfer, M., 2005, Learner Modeling on the

Semantic Web, In Workshop on Personalisation on

the Semantic Web, Edinburgh, Scotland.

Falconer, S. M., Noy, N. F. & Storey, M. A., 2007,

Ontology Mapping - A User Survey, In 2

International Workshop on Ontology Matching at the

International Semantic Web Conference, Busan,

Korea.

Heckmann, D., Schwartz, T., Brandherm, B. & Kröner,

A., 2005, Decentralized User Modeling with UserML

and GUMO, In Workshop on Decentralized, Agent

Based and Social Approaches to User Modelling,

Edinburgh, Scotland.

IMS 2010, IMS Learner Information Package,

http://www.imsglobal.org/profiles.

Kalfoglou, Y. & Schorlemmer, M., 2003, Ontology

Mapping: The State of the Art, The Knowledge

Engineering Review. vol. 18, no. 01, pp. 1-31.

Moodle 2010, http://moodle.org.

Sakai 2010, Sakai Project, http://sakaiproject.org.

Sheth, A. P. & Larson, J. A. 1990, Federated Database

Systems for Managing Distributed, Heterogeneous,

and Autonomous Databases, ACM Computing

Surveys, vol. 22, no. 3, pp. 183-236.

Van Der Sluijs, K. & Houben, G. J., 2006, A generic

component for exchanging user models between web-

based systems, International Journal of Continuing

Engineering Education and Life Long Learning, vol.

16, no. 1, pp. 64-76.

Vassileva, J., McCalla, G. & Greer, J., 2003, Multi-Agent

Multi-User Modeling in I-Help, User Modeling and

User-Adapted Interaction. vol. 13, no. 1, pp. 179-210.

SUPPORTING LEARNER MODEL EXCHANGE IN EDUCATIONAL WEB SYSTEMS

351