Smart Learning Management System Framework

Yeong-Tae Song

, Yuanqiong Wang

, Sungchul Hong

and Yong-Ik Yoon

Towson University, Dept. of Computer and Information Sciences, Towson, MD, U.S.A.

Sookmyung University, Seoul, Korea

Keywords: Smart Learning Management System, Learning Objects, User Profile, Ontology, RDF, Semantic Search.

Abstract Thanks to modern networking technologies and advancement of social networks, people in the modern

society need more and more information just to be in the game. With such environment, the importance of

learning and information sharing cannot be overemphasized. Even though plethora of information is

available on various sources such as the web, libraries, and any learning material repositories, if it is not

readily available and meets the needs of the user, it may not be utilized. For that, we need a system that can

help provide customized information – matches with user’s level and interest - to the user. Such system

should understand what the user’s interests are, what level the user belongs for the topic, and so on. In this

paper, we are proposing a framework for smart learning management system (SLMS) that utilizes user

profiles and semantically organized learning objects so only the relevant information can be delivered to the

user. The SLMS maintains user profiles – continuously updating whenever there is a change – and learning

objects that are organized by building ontology. Upon user’s request, the system fetches relevant learning

materials based on the user’s profile. The delivered learning materials are suitable for the user’s topic and

the level for the requested topic sorted by relevancy ranking.

1 INTRODUCTION

We are living in a highly connected society thanks to

the modern networking technologies and the

Internet. Accessing right information at right time in

the society is not only an essential part of everyday

lives but also considered a main success factor.

Computers are now become a commodity to our

work environment as well as to households. In

WWW alone, there is more than necessary

information available – billions of documents – for

any topic. In addition to that, more information is

available from other sources such as libraries –

digital and traditional - and data repositories – open

to public or available only to closed community. To

be useful, the available information must be directly

related to what users need.

When it comes to e-learning environment, not all

documents are eligible for learning. We can classify

those eligible for learning as learning objects. When

a learner searches for the topic s/he is interested in,

those learning objects should be searchable to meet

the needs of the learner. The search should be tuned

into the learner’s intention and aligned with their

knowledge level. For that need, the learning objects

should be organized into a database using ontology

so that semantic search is possible. For accurate

result, the search should be based on learner’s

profile. Search result then can be ranked by the

relevancy by referencing user’s profile and the

current search topic. It can also be considered as

filtering process so the learning management system

can recommend directly related learning objects

only.

In this paper, we are proposing a smart learning

management system framework where user profile is

maintained dynamically by keeping track of the

changes in the profile. The changes can be either

static, typically provided by user, or dynamic,

typically provided by some type of agent in the

system. Whenever a user is searching for a learning

object, the profile is updated automatically to reflect

the preference and topic of interest. In the

framework, the learning objects are preprocessed –

by building ontology - and stored in the learning

object repository. The preprocessed learning object

information in RDF (Resource Description

Framework) can be used for semantic search. After

each search, the result can be prioritized and indexed

so it can be used for the recommendation to the

learners. The rest of the paper is organized as

follows:



Song Y., Wang Y., Hong S. and Yoon Y..

Smart Learning Management System Framework.

DOI: 10.5220/0004083102290234

In Proceedings of the International Conference on Data Technologies and Applications (DATA-2012), pages 229-234

ISBN: 978-989-8565-18-1

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

Section 2 discusses about the related work,

section 3 discusses learning profile, section 4

discusses learning objects, section 5 illustrates the

smart learning management framework, and section

6 concludes the paper.

2 RELATED WORK

There have been several approaches in making

intelligent learning management system. One

approach is to provide only the relevant learning

materials to the learners. Zaina et al., 2010)

suggested filtering method in preparing learning

objects by referencing user preferences. Ochoa and

Duval (Ochoa and Duval, 2006) proposed contextual

attention metadata for ranking and recommending

learning objects. Another approach is to build

ontology for the learning objects. Keleberda

(Keleberda et al., 2006) proposed a methodology for

building learning object’s and learner’s ontology

using IDEF5 and OWL. The reusability of learning

object in multi-granularity was proposed by Meyer

et al. (Meyer et al., 2011).

2.1 Learner Profile

There are two outstanding learner profile standards:

one is the IMS LIP (Learner Information Package)

standard and the other is the IEEE PAPI (Personal

and Private Information) standard. The PAPI

standard focuses on the tracking of learner’s learning

performance. PAPI’s core data structure elements

are personal information set, relationship to other

users, learner’s security credentials, preference,

performance, and portfolio (Chatti, Klamma, Quix,

& Kensche, 2005; IEEE P1484.2/D8, 2002). LIP is

an XML based structured information model. It

provides rich structure for leaner’s features like goal

and interests. LIP’s meta-data includes time-related

data, identification and indexing information, and

privacy and data protection information. The LIP’s

core data structures are identification, leaning goal,

QCL (Qualification, certification, and licenses),

learning-related activity, transcript, interest,

competency, Affiliation, Accessibility (disability),

Security key, and relationship (set of relationships

between the core components) (IMS, n.d.). In this

paper LIP will be used as a main learner profile

standard and some features of PAPI (e.g.

performance) will be used in the dynamic learner

profile.

2.2 Applicable e-Learning Metadata

Standards

IMS Global Consortium Common Cartridge (IMS

GLC, 2011), IEEE Learning Object Metadata

(LOM) (IEEE ITSC, 2002), and ADL SCORM

(Jesukiewicz, 2009) are the most well-known

metadata standards for describing instructional

resources in e-learning. The main purpose of these

standards is to allow the interoperability for the

learning materials over different learning

management systems.

IMS Global Consortium Common Cartridge

(IMS GLC, 2011) defines a set of open standards

specified in XML, including a format for exchange

of content between systems (Common Cartridge) to

interpret what the digital learning content is and how

it is organized. It is described in a manifest; a

standard for the metadata describing the content in

the cartridge (Learning Object Metadata); a standard

for test items, tests, and assessment (Question and

Test Interoperability) which allows the inclusion of a

question bank; a standard for launching and

exchanging data with external applications (Basic

Learning Tools Interoperability); a controlled

vocabulary to designate the intended use of web

content in the cartridge; a schema for populating

online discussion forums for collaboration among

students; and a schema for populating web links.

IEEE LOM Standard defines a learning object

as any entity -digital or non-digital- that may be used

for learning, education or training. Each metadata

instance describes relevant characteristics of the

learning object to which it applies. The

characteristics are grouped in general (learning

object as a whole), life cycle (history and current

state), meta-metadata (information about the

metadata instance itself), educational (educational

and pedagogic characteristics), technical (technical

requirements and technical characteristics), rights

(intellectual property rights and conditions), relation

(relationship between the learning objects),

annotation (comments on the educational use, when

and by whom the comments were created), and

classification (in relation to a particular

classification system) categories. The LOM data

model is a hierarchy of data elements, including

aggregate data elements and simple data elements.

ADL Sharable Content Object Reference

Model (SCORM) (Jesukiewicz, 2009) is a

collection and harmonization of specifications and

standards that defines the interrelationship of content

objects, data models and protocols such that objects

are sharable across systems that conform to the same

AA   nternational onference on ata echnoloies and Applications



model. This specification promotes reusability and

interoperability of learning content across Learning

Management Systems (LMSs).

3 LEARNER PROFILE

In the proposed smart learning management system

framework, learner profile consists of learner

information and dynamic learning profile.

Learner information is constructed by following

IMS LIP (IMS, n.d.) that defines core data structures

but extended with more detail attributes.

In the competency attribute in the LIP, we have

included desired competency levels (expert, good,

fair, basic, minimal) in addition to what it defines so

the learner can set the learning goal for the desired

topic. Accessibility attribute defines learning

preferences:

 Cognitive preference: e.g. issue of learning

style

 physical preference: e.g. font size

 Technical preferences: e.g. specific computer

platform

Dynamic learning profile is defined as:

 Learner’s performance: frequency and duration

of using learning objects, exam score, etc

 Bookmarks: associated with a topic of interest

 Topics of interest: collected when learners are

using the learning management system

 Learner’s level for the topic

Dynamic information is collected and stored in

the user profile database during the learning process.

4 LEARNING OBJECTS

The purpose of learning objects in e-learning

environment is to deliver learning contents in digital

formats so they can be used in learning management

systems. There are numerous learning objects

available in many different formats and scattered in

many educational organizations or in each

individual’s personal libraries. For efficient use of

learning objects, it must be organized by using

descriptive metadata. The learning objects can be

categorized by general, technical, and educational.

They can also be categorized by the learner’s

preference such as perception, presentation format,

and student participation (Zaina et al., 2010). Since

there are numerous learning objects available, it

should be possible to reuse those objects

systematically when creating new topics. One such

standard is SCORM (sharable content object

reference model) and there is an attempt to reuse

SCORM compliant learning objects in different

granularity (Meyer et al., 2011). In order for those

objects to be reusable, they must be stored in so

called learning object repositories (LORs). A LOR is

a digital archive where users can upload, search and

download learning objects. There are some open

LORs such as MERLOT (MERLOT, n.d.) and

closed LORs – available only to the associated

community - such as Ariadne repository (Duval et

al., 2001).

Ontology is used for organizing learning objects

in this framework. Ontology specifies the

conceptualization of a specific domain in terms of

concepts, attributes and relationship (Noy &

McGuinness, 2001). In general, it defines the

vocabulary and the semantic interconnections and

some simple rules of inference and logic for some

particular topic. It enables the sharing of common

understanding of the structure of information among

people or software agents and the reuse of domain

knowledge. Therefore, it is critical for allowing the

representation, processing, sharing and reuse of the

knowledge among applications in web-based e-

learning systems.

RDF(S) and XML are standards for expressing

ontology in order for it to be shared and reused

(Ghaleb et al., 2006). OWL (Web Ontology

Language) is W3C recommended language for

representing ontologies. OWL is a set of XML

elements and attributes with well-defined meaning.

RDF is a framework that represents metadata,

and a model for representing data about "things on

the Web". It includes a set of triples (subject,

predicate, object). Alternatively, a RDF model can

be represented with a directed labeled graph, or

using an XML-based encoding.

RDF Schema (RDFS) defines the vocabulary of

an RDF model. It uses basic modeling primitives

such as class, subclass-of, property, subproperty-of,

domain, range, and type. RDFS provides

information about the ways in which we describe

our data.

Based on the IEEE LOM standard, we will be

focusing on general and technical perspectives of the

learning object. Learning object ontology will be

used to organize learning objects for easy retrieval.

In our model, specific domain (or learning area)

is defined in a content ontology, while the technical

aspect of the presentation is presented in a structural

ontology. Figure 1 is a snapshot of a sample

ontology created using Protégé 4.1 (Stanford, 2012).

It illustrates how the learning objects can be

mart earnin Manaement ystem rameork



organized. The content ontology is used for

describing the domain structure. In addition to the

definition of the classes in the ontology, the

properties for each class also need to be identified.

For example, the difficulty level of the learning

object will be defined as five levels from "very easy"

to "very difficult" according to the IEEE LOM

standard. The relationships among different topics in

one domain will also be specified. The structural

ontology is used for describing technical details such

as types of activities and how the learning materials

will be presented. Organizing the materials with

these two ontologies make it easier to expand

learning materials into other domains. When

learning materials related to a new domain are

created, only the content ontology needs to be

created for the new domain. When the search is

conducted, the content ontology will be used to

interpret and help identify the relevant learning

object in the repository. The structural ontology will

then be used to identify relevant presentation of the

content (e.g. based on user's profile, a video with

caption is needed to present lecture notes).

Figure 1: A snapshot of sample structural ontology using

Protégé 4.1.

In the figure 2, three nodes are highlighted to

represent one sub-node (behavioral model) being

related to two super nodes (system modeling and

object-oriented analysis and design).

5 THE SMART LEARNING

MANAGEMENT SYSTEM

(SLMS) FRAMEWORK

5.1 Dynamic User Profiler

User profile maintains characteristics of the learner,

which can be categorized into two groups – static

and dynamic. Static information contains learner

information defined by IMS LIP such as goals,

learning preferences, accessibility, etc. and extended

to include more attributes. Dynamic information

includes learning performance, bookmarks, topic of

interest, and user level. Dynamic profiler generates

user profile dynamically from the user profile

database whenever there is a change so the resulting

profile can be up to date anytime.

Figure 2: Snapshot of the content ontology on software

engineering.

5.2 Dynamic User Profile

User profile is an XML document that contains up to

date information about the user’s learning

environment and characteristics of the user. It is

dynamically created whenever there is a change or

new information about the profile.

Figure 3: The SLMS Framework.

5.3 User Level Assessor

User level on the topic will be determined by user

level assessor based on the user profile and the

current topic of interest.

3. User level

assessor

4. Learning

object fetcher

5. Ranking

Learning objects suitable

for the topic and User level

User level for the topic

Topic of

interest

Learner information

(IMS LIP)

Learning

object

Repository

recommended

learning objects

1. Dynamic User

Profiler

Semantic DB

Learning

Object

Learning

Object

Dynamic User Profile

 Learning Performance,

 Bookmarks,

 Topics of interest,

6. learning object

organizer

2. Dynamic

user profile

User

Profile

Learning Content

Management System (LCMS)

AA   nternational onference on ata echnoloies and Applications



5.4 Learning Object Fetcher

Learning object fetcher references user’s dynamic

profile and searches semantic database that is

constructed by the ontology on the learning objects.

The Learning Object Fetch Module (LOFM)

matches the user’s learning goal which is captured in

the learner profiles and the learning objects (LOs).

LOs are stored at Learning Object Repository

according to the content ontology. The meta-data

and the content of these LOs are stored in the form

of RDF. The Learning Object Repository has a

hierarchical structure T, which can have sub-

structures T

, where

T =







(1)

Each T

can be defined by its own sub-structures

recursively. The dynamic user profiler (module 1 in

Figure 3) generates user learning goal, G, in a

hierarchical form. Then the LOFM tries to match G

in T. If a sub-structure matches, let’s say it is T

then the module fetches LOs under T

. The example

of a fetch scenario is shown in the figure 4.

Learning Object Indexing

Learning Object Repository indexing has

multiple levels: the highest level describes the class

hierarchy of an application, e.g. SCORM, the second

level describes attributes and content values, and the

last level describes the structural indices for the

Learning Objects (Chen, Kashyap, & Ghafoor,

2000).

Figure 4: Sample fetch scenario with learning goal and a

part of repository structure.

5.5 Ranking

The ranking module (module 5 in the figure 3)

references dynamic user profile and determines the

relevancy of each learning material in the result. The

result is sorted by relevancy. Once a Learning Goal

is used for learning object (LO) fetching, multiple

LOs could be selected. These Learning Objects may

exist in different sub-trees, which are related by the

content ontology, let’s say sub-tree Tk and Tl are

fetched, where Tk and Tl are the sub-trees of

Learning Object Repository T. At this point the

ranking module evaluates the level of relevancy

between the learning goal G and fetched LOs (Tk

and Tl). If the evaluated relevancy level of Tk is

higher than that of Tl then Tk will receive a higher

ranking. For example, a leaner’s learning goal is

“Behavioral System Modeling”. The Learning

Object fetcher selects sub-trees for behavioral

models under the system modeling and under the

Object Oriented Analysis and Design due to the

relationship between the OOAD and the behavioral

design methodology. Among these two results, the

ranking module will evaluate the system modeling

as higher relevancy to behavioral modeling over

OOAD because of its closer relationship to the

leaning goal – behavioral system modeling.

5.6 Learning Object Organizer

The learning object organizer (module 6 in the

figure 3) constructs ontology by referencing learning

objects. Through the RDF, learning objects have

relations with each other. Section 4.1 discusses how

to build the ontology using software engineering

example.

The snapshot of the prototype of the smart

learning management is in the figure 5. The

prototype is named as “Smart e-Learning Using N-

Screens” or SELNUNS.

Figure 5: SLMS prototype.

6 CONCLUSIONS

Previous literature reported attempts on defining

framework for agent based e-learning systems(e.g.

Rosmalen et al., 2005; Zaina et al., 2010) However,

most of the framework reported did not provide

details on issues such as how user profile can be

establish and maintained, how the LOs can be

organized, and how the semantic search can be

conducted in order to present LOs that are the most

relevant to the learners needs (including content,

Learning Goal G

Learning Object Repository T

mart earnin Manaement ystem rameork



learning activities, and presentation styles, etc).

In this paper, we have proposed a framework for

smart learning management system. The framework

is designed to provide learning materials that is

suitable for the user level for the topic and most

relevant to the user’s interest. To improve the

accuracy, we have separated user profile into two

parts – static and dynamic, which is an extension of

IMS LIP definition. Dynamic profiler maintains up

to date information for user profile so the learning

object fetcher and the user level assessor can

reference them to provide more accurate result to the

learner. Learning objects are organized by

constructing ontology. Based on RDF description,

semantic search for the learning objects can be

conducted. The final results are filtered by using

relevancy ranking and the filtered result is delivered

to the learner as recommended learning objects.

ACKNOWLEDGEMENTS

This research work is funded by Korea Association

of Industry, Academy, and Research Institute

(KAIRI).

REFERENCES

Chatti, M. A., Klamma, R., Quix, C., & Kensche, D.

(2005). LM-DTM: an environment for XML-based,

LIP/PAPI-compliant deployment, transformation and

matching of learner models. Fifth IEEE International

Conference on Advanced Learning Technologies, 567-

569. Kaosiung, Taiwan.

Chen, S.-C., Kashyap, R. L., & Ghafoor, A. (2000).

Semantic models for multimedia database searching

and browsing. Boston, MA: Kluwer Academic

Publishers.

Duval, E., Forte, E., Cardinaels, K., Verhoeven, B., Van

Durm, R., Hendrikx, K., Forte, M. W., et al. (2001).

The Ariadne knowledge pool system. Communications

of the ACM, 44(5), 72-78.

Ghaleb, F., Daoud, S., Hasna, A., Alja, J. M., El-seoud, S.

A., & El-sofany, H. (2006). E-Learning Model Based

On Semantic Web Technology. International Journal

of Computing & Information Sciences, 4(2), 63-71.

IEEE P1484.2/D8. (2002). Public and Private Information

(PAPI) for Learners. Retrieved from http://edutools.

com/papi/

IMS. (n.d.). IMS Learner Information Packaging

Information Model Specification. Retrieved March 20,

2012, from http://www.imsglobal.org/profiles/lipinfo

01.html

IMS GLC. (2011). IMS GLC Common Cartridge Profile:

Overview. Retrieved from http://www.imsglobal.org/

cc/ccv1p2/imscc_profilev1p2-Overview.html

Jesukiewicz, P. (2009). SCORM 2004 Content Aggrega-

tion Model. Learning. Retrieved from http://www.adl

net.gov/capabilities/scorm/scorm-2004-4

#tab-resour

ces

Keleberda, I., Repka, V., & Biletskiy, Y. (2006). Building

Learner’s Ontologies to Assist Personalized Search of

Learning Objects. In ICEC’06, 569-573. Fredericton,

Canda.

MERLOT. (n.d.). MERLOT - Multimedia Educational

Resource for Learning and Online Teaching. Retrieved

March 20, 2012, from http://www.merlot.org/merlot/

index.htm

Meyer, M., Rensing, C., & Steinmetz, R. (2011).

Multigranularity reuse of learning resources. ACM

Transactions on Multimedia Computing, Communi-

cations, and Applications, 7(1), 1-23.

Noy, N. F., & McGuinness, D. L. (2001). Ontology

development 101: A guide to creating your first

ontology. Development. Stanford knowledge systems

laboratory technical report KSL-01-05 and Stanford

medical informatics technical report SMI-2001-0880.

Retrieved from http://er.uni-koblenz.de/IFI/AGStaab/

Teaching/SS09/sw09/Ontology101.pdf

Ochoa, X., & Duval, E. (2006). Use of contextualized

attention metadata for ranking and recommending

learning objects. In CAMA ’06, 9-15. ACM Press.

Rosmalen, P. V., Brouns, F., Tattersall, C., Vogten, H.,

Bruggen, J. V., & Koper, R. (2005). Towards an open

framework for adaptive, agent-supported e-learning.

International Journal of Continuing Engineering

Education and Life Long Learning, 15(3), 261-275.

Retrieved from http://dspace.learningnetworks.org/

retrieve/94/

Stanford. (2012). Protégé. Retrieved from http://protege.

stanford.edu/

Zaina, L. A. M., Rodrigues, J. F., & Bressan, G. (2010).

An approach to design the student interaction based on

the recommendation of e-learning objects. In

SIGDOC'10, the 28th ACM International Conference

on Design of Communication, 223. ACM Press.

AA   nternational onference on ata echnoloies and Applications

