A novel method for generating an elearning ontology

International Journal of Data Mining & Knowledge Management Process (IJDKP) Vol.3, No.6, November 2013

A NOVEL METHOD FOR GENERATING AN ELEARNING ONTOLOGY
Mohammed Reda Chbihi Louhdi1, Hicham Behja2 and Said Ouatik El Alaoui3
1,3

Laboratoire Informatique & Modélisation,
Faculty of sciences Dhar El Mehraz, Fez, Morocco
2
Ecole Nationale Supérieure d'Electricité et de Mécanique, Casablanca, Morocco

ABSTRACT
The Semantic Web provides a common framework that allows data to be shared and reused across
applications, enterprises, and community boundaries. The existing web applications need to express
semantics that can be extracted from users' navigation and content, in order to fulfill users' needs. Elearning has specific requirements that can be satisfied through the extraction of semantics from learning
management systems (LMS) that use relational databases (RDB) as backend. In this paper, we propose
transformation rules for building owl ontology from the RDB of the open source LMS Moodle. It allows
transforming all possible cases in RDBs into ontological constructs. The proposed rules are enriched by
analyzing stored data to detect disjointness and totalness constraints in hierarchies, and calculating the
participation level of tables in n-ary relations. In addition, our technique is generic; hence it can be applied
to any RDB.

KEYWORDS
Semantic Web, Ontology Building, E-learning, Learning Management System, Relational Databases,
Schema Mapping, Data Analysis.

1. INTRODUCTION
Nowadays, the Web became a huge universe in perpetual expansion. It's consists of interlinked
pages and web applications that contains a lot of videos, photos and interactive contents. Over
time web technologies have evolved to give web developers the ability to create new generations
of useful and immersive web experiences. Today's web is a result of the ongoing efforts of an
open web community that helps define these web technologies, like HTML5, CSS3 and WebGL
and ensure that they're supported in all web browsers.
Before 1999, experts call the Internet "Read-Only" web. The average internet user’s role was
limited to reading the information which was presented to him. The lack of active interaction of
common users with the web leads to the birth of Web 2.0 or the "read-write" web. It has the
ability to contribute content and interact with other web users. This interaction and contribution
has dramatically changed the landscape of the web, and created billions of terabytes of data. To
manage this data, it must be machine understandable. The web 3.0 or the "semantic web"
provides a common framework that allows data to be shared and reused across applications,
enterprises, and community boundaries. One of its objectives is to convert the current web into a
"web of data", by encouraging the inclusion of semantic content in web pages and documents.
Besides, the Semantic Web aims at making information on the Web machine processable and
understandable, and therefore, facilitates interoperability between applications [1].
DOI : 10.5121/ijdkp.2013.3610

151


Currently, most of the existing web applications in multiple domains (e-learning, e-government,
e-commerce, etc.) create and manage data in structured or semi-structured way without
expressing semantics. This semantic can help enhancing the structure and content of the web
sites. It can also be used by recommendation systems to provide users with suggestions to meet
their informational needs and preferences. Semantics can be extracted from two sources:
-

The users ' navigation in order to better understand what users need,
The content, which is generally structured in relational databases, in order to make it
machine understandable,

Semantics are expressed through a technology called ontology, which is a formal and explicit
specification of a shared conceptualization that refers to an abstract model of some phenomenon
in the world that identifies the relevant concepts of that phenomenon [2]. Ontologies are
particularly specified in languages that make possible the abstraction of data structures and allow
strategies implementation. Semantic Web is then expected to provide languages that can both
express data and rules for reasoning about the data, and also to export rules from any existing
knowledge-representation system onto the web.
E-learning is just-in-time education integrated with high velocity value chains. It is the delivery of
individualized, comprehensive, dynamic learning content in real time, aiding the development of
communities of knowledge, linking learners and practitioners with experts [3]. The web 2.0
provides new tools (like forums and wikis) that enhances the e-learning and upgrades it to the so
called "collaborative e-learning" or the "e-learning 2.0" [4]. Actually, new needs appear, the
learning management systems (LMS) must provide a custom content to the learners, according to
their preferences, experiences and needs. To reach this objective, semantics must be extracted
from the LMS (user's navigation and content).
In this work we will focus on extracting semantics from the content of an open source LMS:
"Moodle". It uses a relational database as a backend to manage its data. This database
encapsulates conceptual data models and hides a strong semantics that could be exploited in the
process of ontology extraction. Thus, we propose a method that uses schema mapping technique
combined with a data analysis to generate an OWL ontology from a relational database. Our
proposal covers all possible cases in databases and allows having richer ontologies.
The rest of this paper is organized as follows. Section 2 discusses related works in
recommendation systems for Learning Management Systems, and the methods for extracting
semantics using ontology engineering from relational databases. Section 3 describes the proposed
transformation rules. Implementation in Moodle is presented in Section 4. Finally, Section 5
concludes this paper, and discusses the perspectives of this work.

2. RELATED WORKS
Due to the exponential increase in the amount of resources available and accessible on the web,
Recommendation systems have seen their popularity grow in recent years. Combining techniques
of information filtering, personalization, artificial intelligence, social networks and humancomputer interaction, recommendation systems provide users with suggestions to meet their
informational needs and preferences [5].
To produce recommendations, a number of approaches is possible: (1) the approach by content
[6] which makes recommendations by comparing the semantic content of resources with the
user’s tastes, (2) the approach based on knowledge [7] that makes recommendations by exploiting
knowledge about the user and pre-established heuristics, and (3) the approach by collaborative
152


filtering [8], which makes recommendations by analyzing, at the same time, the user’s opinions
and those of other users about the resources they have consulted.
Many works proposed to integrate recommendation systems in Learning Management Systems,
in order to meet the specific needs of each learner. In [5] authors propose a Plugin that can be
integrated on any e-Learning platform to keep tracks of the web searches made by learners, in
order to use it for the recommendation. This Plugin also includes a new hybrid method to rank
web documents before presenting them to the learner. In the paper [9], authors describe an
automatic personalization approach aiming to provide online automatic recommendations for
active learners without requiring their explicit feedback. Recommended learning resources are
computed based on the current learner’s recent navigation history, as well as exploiting
similarities and dissimilarities among learners’ preferences and educational content. A technique
for developing a solution of personalized recommendations for e-learning systems is presented in
[10]. It adopts an ontology-based modeling of user profiles and document models. The proposed
recommendation technique is a hybrid one that involves two phases: (1) Collaborative filtering
phase: user conceptual navigation is tracked in order to predict the next concept which will be
focused by the user, according to his fingerprints and interests profile, (2) Content-based filtering:
this concept is used in order to select the documents to be effectively recommended, in
concordance with the user competences profile. In [11], the user-based collaborative filtering
method is chosen as the primary recommendation algorithm, combined with online education.
The authors analyze the requirement of a web-based e-learning recommendation system, and
divide the system workflow into five sections: data collection, data ETL, model generation,
strategy configuration, and service supply. Moreover, an architecture is proposed, based on which
further development can be accomplished. In this architecture, there are seven modules, and four
of them are core modules: recommendation models database, recommendation system database,
recommendation management, data/model management.
Most of Learning Management Systems use relational databases as backend. Those databases
contain strong semantics that can be exploited in recommendation systems and store important
and useful information, which is a valuable source for ontology learning. The existing methods
for ontology engineering from relational databases are based on three types of techniques [12],
[13]:
-

Reverse engineering: consists in recovering the conceptual data model from the physical
schema of an existing database [14],
Schema mapping: transforms the components of the conceptual data model or the
physical model into ontology's concepts and relations [12],
Data Mining [15]: exploits the records of the database in order to extract knowledge.

Recently, some methods [16], [17], [18], [19] proposed to combine schema mapping with reverse
engineering [16], [18] or with Data Mining [17], [19]. Their objective is to extract more
information from the database that will make richer the generated ontology. Methods that focus
merely on transforming database schema generate inconsistent ontologies having a weak
semantics. In other words, simple schema mapping leads to weak semantics and inconsistency
and hence, will not be able to extract information enabling strong ontology building, especially in
case of highly abstracted databases.
However, some existing methods [16], [18], [20], [21], [22] use the conceptual data model as a
source of ontology learning, because it is semantically richer than the relational model.
Unfortunately, in the most cases, the databases are available in a physical format (the
corresponding conceptual data model is not available). In addition, the mapping operation (from
the conceptual data model to the relational model) may create new tables and new attributes that
153


make difficult the ontology populating task (since there are many differences between the
components of the conceptual data model and those of the relational model).
Schema mapping technique is used by ontology building methods. It converts the relational
database schema (or the ER Model) to an ontology by using a set of predefined transformation
rules. Some works [23], [24], [25], [26], [27] map ontologies to relational databases schemata in
order to maintain interoperability between them. The schema mapping is performed using
mapping rules that update the ontology when the database is modified and vice-versa.
Note that most methods based on the schema mapping technique cannot handle some complex
cases like multiple inheritance, many-to-many relations with attributes, and the n-ary relations.
The multiple inheritance case was treated by [22] where authors reproduce the hierarchy found in
the conceptual data model in the taxonomy of the ontology. Concerning the many-to-many
relations with attributes, they were supported by the transformation rules of some methods [20],
[21], [22], [28]. The n-ary relation is a difficult case, because only binary relations between
classes can be represented through object properties in the ontology. However, some works [20],
[21], [28], [29], [30] propose solutions to represent n-ary relations in OWL ontologies. In [21],
[28], [29], [30], authors create a class for the bridge table related by two object properties
mutually inverse. The method [20] uses AllValuesFrom restrictions to link the class
corresponding to the bridge table, with the classes that correspond to the participating tables to
the n-ary relation. This solution is more representative than the first one, because the existence of
a record in the bridge table is conditioned by the existence of records in tables that participate to
the n-ary relation.
The foreign key columns (or one-to-many relations) and the simple inheritance cases are
processed by the most of existing methods. Furthermore, the most existing methods transform the
simple attributes into Data Type Properties. Some of them [21], [22], [28], [29], [30] suggest to
add restrictions to the attributes that have a constraint (Primary Key, NOT NULL or UNIQUE).
In the Table I, we present the main methods and the different cases treated from the relational
model.
Table 1. The main methods and the different cases captured from the relational model.

Methods
[16], [17]
[18], [19],
[31]
[20], [28]
[21]
[22]
[29], [30]
[32], [33],
[34], [35],
[36]
Our method

One-tomany
relation

Simple
inheritance

X

Multiple
inheritance

X

X

Many-tomany
relation

Many-tomany
relation with
attributes

n-ary
relations

X
X
X

X
X

X

X
X
X
X

X

X

X

X

X

X
X

X

X
X
X
X

X

X
X

X

X

X

In this work, we propose to use the relational model as a source for ontology learning. Unlike the
methods mentioned previously, we propose exhaustive transformation rules that deal with most of
existing cases in databases (table1). Moreover, we analyze the database records to recover some
disappeared aspects during the mapping from the conceptual data model to the relational model
154


(like disjointness and totalness in simple inheritance cases and the participating level of tables in
n-ary relations).

3. TRANSFORMATION RULES
The process of transforming a relational database into an OWL ontology begin with classifying
tables of the database schema into six categories according to their attributes. After that, we
transform the tables of each category into ontological components by applying both mapping
rules and data analysis. The latter finds disjointness and totalness in inheritance cases, and spots
the participation level of tables in n-ary relations. The table 2 shows the proposed classification.
In this work, we suppose that the databases are at least in the third normal form. Note that the first
two categories contains only strong entities, hence the four others contains weak entities.
Table 2. The different categories adopted for classifying the database tables.
Category
1
2
3
4

5

6

Features
Tables containing only simple attributes without foreign keys constraint
(Example : Tables PERSON and PROGRAM in Figure 1)
Tables containing at least one foreign key (Example: Table ACTIVITY in Figure
1).
Tables whose entire primary key is also a foreign key referencing a single table.
(Example: Tables STUDENT and TEACHER in Figure 1).
Tables containing a composite primary key (two or more fields) which is also a
foreign key whose fields are referencing exactly two tables (Example: Table
SUPERVISION in Figure 1).
foreign key whose fields are referencing more than two tables. Simple attributes
are not duplicated in any of the referenced tables (Example: Table
OFFERED_COURSE in Figure 1).
foreign key whose fields are referencing more than two tables. Some simple
attributes are duplicated in the referenced tables (Example: Table
PEDAGOGICAL_PROJECT in Figure 1).

After classifying the database tables, we apply the appropriate transformation rules for each
table's category (see table 2). In the rest of this paragraph, we will present the proposed
transformation rules. All these rules are illustrated by examples using the database which is
presented in Figure 1.
Rule 1: The tables that contain only simple columns (without foreign key constraint) are
transformed into simple classes into the ontology (category 1). Example:
<owl:Class rdf:ID="PERSON"/>
<owl:Class rdf:ID="PROGRAM"/>

155

PERSON
Ssn
Name

BirthDate

Address

EMPLOYEE
#Ssn
Salary

STUDENT
#Ssn #ProgramID

ENGINEER
#Ssn EngType

SECRETARY
#Ssn TypingSpeed

TECHNICIAN
#Ssn Tgrade

TEACHER
#Ssn Speciality

PROGRAM
ProgramID

SUPERVISION
#Ssn
#ProjectID
PROJECT
ProjectID

ProjectName

ACTIVITY
ActivityID

ActivityName

OFFRED_COURSE
#Ssn #ProgramID

Duration

#ProjectID
Project_ID

PEDAGOGICAL_PROJECT
#CourseID
#ProjectID ProjectName

#CourseID

COURSE
CourseID

Duration

ProgramName

CourseName

CourseName

LearningHours

LearningHours

Figure 1. Example of a normalized database

Rule 2: Tables of the second category are transformed into simple classes in the ontology. Each
foreign key is mapped into two Object-Properties (mutually inverse). The first one has the class
corresponding to current table as domain, and its range is the referenced table by the foreign key.
The second one (inverse of the first Object-Property) is declared as inverse functional. Example:
<owl:Class rdf:ID="ACTIVITY"/>
<owl:ObjectProperty rdf:ID="activityHasProject">
<rdfs:domain rdf:resource="#ACTIVITY" />
<rdfs:range rdf:resource="#PROJECT" />
</owl:ObjectProperty>
<owl:InverseFunctionalProperty rdf:ID="project'sActivity">
<rdfs:domain rdf:resource="#PROJECT" />
<rdfs:range rdf:resource="#ACTIVITY"/>
<owl:inverseOf rdf:resource="#activityHasProject" />
</owl:InverseFunctionalProperty>

Rule 3: we can identify simple inheritance relationships from tables of the third category. All
tables in this category are sub-tables in hierarchies. Each sub-table is transformed into a class in
the ontology and is declared as a subclass of the table referenced by the foreign key (which is also
the primary key of each sub-table). Example:
<owl:Class rdf:ID="PERSON" />
<owl:Class rdf:ID="STUDENT">
<rdfs:subClassOf rdf:resource="#PERSON" />
</owl:Class>
<owl:Class rdf:ID="TEACHER"/>
<rdfs:subClassOf rdf:resource="#PERSON" />
</owl:Class>

After reproducing simple inheritance relations into the taxonomy of the ontology, we will identify
disjointness and totalness constraints in those relations.
156


In a simple inheritance relation, disjointness means that an entity can be a member of at most one
of the subclasses (that have the same level) of a hierarchy [37]. To identify the existence of
disjointness between tables having the same level in a hierarchy, we propose the algorithm
presented in Figure 2.
Disjointness
Input: SC as a list of the sub-tables of a simple inheritance relation
Output: DL as a two-dimensional array that will contain disjoint tables
Let pk(T) a function retrieving the primary key of a table T
Let val(attr) a function that retrieves the value of the attribute "attr" for the current record
Let N the size of the list SC
FOR i = 0 to N-1
FOR j = i+1 to N-1
FOR each record of the table SC[i]
FOR each record of the table SC[j]
IF val(pk(SC[i])) = val(pk(SC[j])) THEN
BREAK
ENDIF
ENDFOR
ENDFOR
Add [SC[i], SC[j]] to the array DL
ENDFOR
ENDFOR
Figure 2. Disjointness detection algorithm

The following OWL code illustrates the disjointness constraint in a simple inheritance relation:
<owl:Class rdf:about="#TECHNICIAN">
<rdfs:subClassOf rdf:resource="#EMPLOYEE"/>
<owl:disjointWith rdf:resource="#SECRETARY"/>
<owl:disjointWith rdf:resource="#TEACHER"/>
</owl:Class>
<owl:Class rdf:about="#SECRETARY">
</owl:Class>
<owl:Class rdf:about="#TEACHER">
</owl:Class>

Concerning the totalness, it specifies that every entity in the superclass must be a member of at
least one subclass in the hierarchy [37]. To identify the existence of totalness in a hierarchy, we
propose the algorithm presented in Figure 3. The following OWL code illustrates the totalness
constraint in a simple inheritance relation:
<owl:class rdf:ID="EMPLOYEE">
<owl:unionOf rdf:parseType="Collection">
<owl:class rdf:about="#SECRETARY" />
<owl:class rdf:about="#TEACHER" />
<owl:class rdf:about="#TECHNICIAN" />
</owl:unionOf>
</owl:class>

157


Totalness
Input: - ST : the super-table
- ssT : a list of the sub-tables of ST
Output: F as Boolean to flag the existence of totalness
Let pk(T) a function retrieving the primary key of a table T
Let val(attr) a function that retrieves the value of the attribute "attr" for the current record
Let N the size of the list ssT
F FALSE
FOR each table T of ssT
FOR each record of ST
FOR each record of T
IF val(pk(T)) = val(pk(ST))THEN
N N–1
BREAK
ENDIF
ENDFOR
ENDFOR
ENDFOR
IF N = 0 THEN
F TRUE
ENDIF
Figure 3. Totalness detection algorithm

In the case of existence of both a disjointness and totalness in a simple inheritance relationship,
we combine the two previous proposals. The following example illustrates this situation:
<owl:Class rdf:about="#TECHNICIAN">
<owl:disjointWith rdf:resource="#SECRETARY"/>
</owl:Class>
<owl:Class rdf:about="#SECRETARY">
</owl:Class>
<owl:Class rdf:about="#TEACHER">
</owl:Class>
<owl:class rdf:ID="EMPLOYEE">
<owl:unionOf rdf :parseType="Collection">
<owl:class rdf:about="#SECRETARY" />
<owl:class rdf:about="#TEACHER" />
<owl:class rdf:about="#TECHNICIAN" />
</owl:unionOf>
</owl:class>

Rule 4: The tables containing a composite primary key (two or more columns) which is also a
foreign key whose fields are referencing exactly two tables (category 4), are mapped into two
Object-Properties mutually inverse. Example:
<owl:ObjectProperty rdf:ID="hasProject">
<rdfs:domain rdf:resource="#ENGINEER" />
<rdfs:range rdf:resource="#PROJECT" />
158


<owl:ObjectProperty rdf:ID="hasEngineer">
<rdfs:domain rdf:resource="#PROJECT" />
<rdfs:range rdf:resource="#ENGINEER" />
<owl:inverseOf rdf:resource="#hasProject"/>

If a table of category 4 contains simple columns (table resulted from a many-to-many relation
with attributes), we apply rules 4 and 5 to this table.
Rule 5: the tables of the category 5 are resulting from n-ary relations. Their primary keys are
composed by several foreign keys (more than two) referencing the participating tables to the
relation.
To represent n-ary relations from the relational model, we create a class corresponding to the
bridge table related to the classes that correspond to the participating tables to the n-ary relation
by OWL restrictions (allValuesFrom or someValuesFrom). These restrictions are depending on
the participation level of tables in the relation. In Figure 6, we present an example of transforming
the n-ary relation OFFRED_COURSE in the example given in the Figure 1.
To define the participation level of each table to the n-ary relation, we check if all records of the
participating tables are referenced in the bridge table. If so, we use an allValuesFrom restriction,
else a someValuesFrom restriction is used.
OFFRED_COURSE

allValuesFrom

someValuesFrom

TEACHER

PROGRAM

someValuesFrom
COURSE

Figure 6. Example of transforming a n-ary relation

To illustrate this solution, we present below the OWL code corresponding to the n-ary relation
OFFRED_COURSE:
<owl:Class rdf:ID="OFFRED_COURSE">
<rdfs:subClassOf>
<owl:Restriction>
<owl:someValuesFrom>
<owl:Class rdf:about="#COURSE"/>
</owl:someValuesFrom>
<owl:onProperty>
<owl:ObjectProperty rdf:about="hasCourse"/>
</owl:onProperty>
</owl:Restriction>
</rdfs:subClassOf>
<rdfs:subClassOf>
<owl:Restriction>
<owl:someValuesFrom>
<owl:Class rdf:about="#PROGRAM"/>
</owl:someValuesFrom>
<owl:onProperty>
<owl:ObjectProperty rdf:about="hasProgram"/>
</owl:onProperty>
</owl:Restriction>
</rdfs:subClassOf>
159


<rdfs:subClassOf>
<owl:Restriction>
<owl:allValuesFrom>
<owl:Class rdf:about="#TEACHER"/>
</owl:allValuesFrom>
<owl:onProperty>
<owl:ObjectProperty rdf:about="hasTeacher"/>
</owl:onProperty>
</owl:Restriction>
</rdfs:subClassOf>
</owl:Class>

Rule 6: in the conceptual data model, the tables of the category 6 are subclasses in more than one
class/subclass relationship (multiple inheritance). In the relational model, these tables can be
confused with the bridge tables of n-ary relations (category 5), since they are weak entities whose
primary key consists of two or more foreign keys referencing two or more tables. To distinguish
between both categories (5 and 6), we supposed that tables of category 6 must contain in addition
to the primary key, inherited attributes (during the mapping process) belonging to super-tables.
To map the multiple inheritance case, we reproduce the same hierarchy in the taxonomy of the
ontology. Each sub-table (of category 6) is transformed into a subclass of the classes
corresponding to the tables referenced by the foreign keys of the sub-table. For example, the
OWL code for the PEDAGOGICAL_PROJECT (Figure 1) is as follows:
<owl:Class rdf:ID="COURSE" />
<owl:Class rdf:ID="PROJECT"/>
<owl:Class rdf:ID="PEDAGOGICAL_PROJECT">
<rdfs:subClassOf rdf:resource="#PROJECT" />
<rdfs:subClassOf rdf:resource="#COURSE" />
</owl:Class>

Rule 7: Concerning the transformation of the columns (without foreign key constraint), we create
for each attribute a dataType property for which the domain is the class corresponding to the table
containing this column and the range is the type in XML schema. Example:
<owl:DatatypeProperty rdf:ID="Name">
<rdfs:domain rdf:resource="#PERSON"/>
<rdf:range rdf:resource="&xsd;String"/>
</owl:DatatypeProperty>

For attributes with special constraints such as NOT NULL, UNIQUE and Primary Key, we
propose to treat them as follows:
NOT NULL: add the MinCardinality restriction to the Datatype Property with the value 1,
<owl:Class rdf:ID="PERSON">
<rdfs:subClassOf>
<owl:Restriction>
<owl:onProperty rdf:resource="#Name" />
<owl:minCardinality
rdf:datatype="&xsd;nonNegativeInteger">1</owl:minCardinality>
</owl:Restriction>
</rdfs:subClassOf>
</owl:Class>

UNIQUE: declare the Datatype property as inverse functional,
160


<owl:InverseFunctionalProperty rdf:ID="Program_name">
<rdfs:domain rdf:resource="#PROGRAM"/>

Primary Key: add the MinCardinality restriction to the Datatype Property with the value 1, and
declare it as inverse functional,
<owl:InverseFunctionalProperty rdf:ID="Ssn">
<rdfs:domain rdf:resource="#PERSON"/>
<owl:Class rdf:ID="PERSON">
<rdfs:subClassOf>
<owl:Restriction>
<owl:onProperty rdf:resource="#Ssn" />
<owl:minCardinality
rdf:datatype="&xsd;nonNegativeInteger">1</owl:minCardinality>
</owl:Restriction>
</rdfs:subClassOf>
</owl:Class>

The table 3 summarizes the proposed transformation rules.
Table 3. Summary of the proposed transformation rules.
Rule
1

Case
Strong entity
Foreign Key

2

3&6
4

Simple and Multiple
inheritance
Many-to-Many
relation
N-ary Relation

5
4&5

7

Many-to-Many
relation with attributes
Simple column

Column with
UNIQUE constraint

OWL Component
Class
Two Object Properties mutually inverse
1. The First one :
- Domain: Class corresponding to the table containing the
column,
- Range : Class corresponding to the referenced table,
2. The Second one is the inverse of the first, and it is declared as
inverse functional
Reproducing inheritance relations into the taxonomy of the ontology
Two Object Properties mutually inverse that rely the classes
corresponding to the participant tables of the relation
A class corresponding to bridge table, related to the classes that
correspond to the participating tables with "is-a" relation restricted
with the OWL restriction "AllValuesFrom" or "SomeValuesFrom"
according to the participating level of each table of the relation
Combination of the above two cases
Data Type Property
- Domain : Class corresponding to the table containing the
column,
- Range : The column type expressed with XML Schema,
Inverse Functional Property
column,
- Range: The column type expressed with XML Schema,

161


Column with NOT
NULL constraint

Primary Key

Data Type Property
column,
- Minimal cardinality = 1
Inverse Functional Property
column,
- Minimal cardinality = 1

4. IMPLEMENTATION IN MOODLE LMS
To evaluate the efficiency of the proposed transformation rules, we implemented the proposal
with Java and the Jena API (Java framework for building Semantic Web applications) for
automatically OWL ontology building from relational database. Having a friendly-user interface,
our system allows building OWL file that contains the definition of the extracted ontology from a
MySql database.
We use the database of the Learning Management System of the Cadi Ayyad University (CAU)
of Marrakech – Morocco (http://minfo.uca.ma/), which is based on the open source LMS
Moodle. This system is used to provide computer sciences courses for students of the bachelor
program. Moreover, collaborative tools (forums, wikis and blogs) are enabled in order to allow
learners and authors to discusses and collaborate.
The Moodle database isn't in normalized form. All the primary keys of all tables have the same
names: "id". After exporting the database of the LMS of CAU, all foreign keys constraints
disappear. To recover them, we use the database creation files (in XML format) that contain the
metadata of all tables.
The exported database contains 198 tables (for the version 1.9.4 of Moodle). Otherwise, the
pedagogical platform of the CAU (in production since 2009) contains 107 courses and more than
39000 users (until July 2013). Some statistics about the stored data are presented in the table 4,
and the metadata of the database structure are presented in the table 5.
Table 4. Some statistics about the pedagogical platform of the CAU.
Category

Users

Information
Administrator
Course Responsible
Teacher Editor
Teacher non Editor
Learner
All users

Forums
Other collaboration tools

Discussions
Topics
Posts
Wikis
Messages échangés en CHAT

Number
5
2
73
157
38874
39111
354
224
4654
4
3373

162


Table 5. Metadata of Moodle database of the pedagogical platform of the CAU.

Entities

Columns

Metadata
Strong
Weak
Primary Key (PK) constraint
Foreign Key (FK) constraint
PK and FK constraints
Others

Count
198
0
198
166
0
1431

Using the proposed transformation rules to map Moodle database into ontology, the obtained
numbers of concepts, Object Properties and Data Type Properties are respectively 198, 332 and
1629. All the existing cases in the Moodle database were successfully transformed into
ontological components. The resulted OWL file is very big (more than 14000 lines). Due to the
number of pages limit, we will present a small part of the ontology that concerns a part of the
Moodle question's module. The relational model, the ontology (presenting the Object Properties
relations between concepts) and the OWL code (containing the concepts, Object Properties and
some Data Type Properties) of the question's module are presented respectively in the figures 7, 8
and 9.

Figure 7. The relational model of a sub-part of the Question's module in the Moodle Database

163


Figure 8. The ontological model of a sub-part of the Question's module in the Moodle Database
<owl:Class rdf:about="question_numerical"/>
<owl:Class rdf:about="question_multichoice"/>
<owl:Class rdf:about="question_answers"/>
<owl:Class rdf:about="question_truefalse"/>
<owl:Class rdf:about="question_shortanswer"/>
<owl:Class rdf:about="question"/>
<owl:ObjectProperty rdf:about="question_multichoiceHasquestion">
<rdfs:range rdf:resource="question"/>
<rdfs:domain rdf:resource="question_multichoice"/>
<owl:ObjectProperty rdf:about="questionHasQuestion">
<rdfs:domain rdf:resource="question"/>
<owl:ObjectProperty rdf:about="question_truefalseHasQuestion">
<rdfs:domain rdf:resource="question_truefalse"/>
<owl:ObjectProperty rdf:about="question_answersHasQuestion">
<rdfs:domain rdf:resource="question_answers"/>
<owl:ObjectProperty rdf:about="question_shortanswerHasQuestion">
<rdfs:domain rdf:resource="question_shortanswer"/>
<owl:ObjectProperty rdf:about="question_numericalHasQuestion">
<rdfs:domain rdf:resource="question_numerical"/>
<owl:InverseFunctionalProperty rdf:about="question'sQuestion_multichoice">

164

<owl:inverseOf rdf:resource="question_multichoiceHasQuestion"/>
<rdfs:range rdf:resource="question_multichoice"/>
<rdf:type rdf:resource="http://www.w3.org/2002/07/owl#ObjectProperty"/>
<owl:InverseFunctionalProperty rdf:about="question'sQuestion">
<owl:inverseOf rdf:resource="questionHasQuestion"/>
<owl:InverseFunctionalProperty rdf:about="question'sQuestion_truefalse">
<owl:inverseOf rdf:resource="question_truefalseHasQuestion"/>
<rdfs:range rdf:resource="question_truefalse"/>
<owl:InverseFunctionalProperty rdf:about="question'sQuestion_answers">
<owl:inverseOf rdf:resource="question_answersHasQuestion"/>
<rdfs:range rdf:resource="question_answers"/>
<owl:InverseFunctionalProperty rdf:about="question'sQuestion_shortanswer">
<owl:inverseOf rdf:resource="question_shortanswerHasQuestion"/>
<rdfs:range rdf:resource="question_shortanswer"/>
<owl:InverseFunctionalProperty rdf:about="question'sQuestion_numerical">
<owl:inverseOf rdf:resource="question_numericalHasQuestion"/>
<rdfs:range rdf:resource="question_numerical"/>
<owl:InverseFunctionalProperty rdf:about="id_question_truefalse">
<rdfs:range rdf:resource="http://www.w3.org/2001/XMLSchema#integer"/>
<rdf:type rdf:resource="http://www.w3.org/2002/07/owl#DatatypeProperty"/>
<owl:InverseFunctionalProperty rdf:about="id_question_multichoice">
<owl:InverseFunctionalProperty rdf:about="id_question_answers">
<owl:InverseFunctionalProperty rdf:about="id_questionS_numerical">
<owl:InverseFunctionalProperty rdf:about="id_question_shortanswer">
<owl:InverseFunctionalProperty rdf:about="id_question">

165

<owl:Restriction>
<owl:minCardinality
rdf:datatype="http://www.w3.org/2001/XMLSchema#int">1</owl:minCardinality>
<owl:onProperty rdf:resource="id_question_truefalse"/>
</owl:Restriction>
<owl:Restriction>
<owl:minCardinality
<owl:onProperty rdf:resource="id_question_multichoice"/>
</owl:Restriction>
<owl:Restriction>
<owl:minCardinality
<owl:onProperty rdf:resource="id_question_answers"/>
</owl:Restriction>
<owl:Restriction>
<owl:minCardinality
<owl:onProperty rdf:resource="id_question_numerical"/>
</owl:Restriction>
<owl:Restriction>
<owl:minCardinality
<owl:onProperty rdf:resource="id_question_shortanswer"/>
</owl:Restriction>
<owl:Restriction>
<owl:minCardinality
<owl:onProperty rdf:resource="id_question"/>
</owl:Restriction>
<owl:DatatypeProperty rdf:about="trueanswer_question_truefalse">
<owl:Restriction>
<owl:minCardinality
<owl:onProperty rdf:resource="question_truefalse"/>
</owl:Restriction>
wl:DatatypeProperty rdf:about="createdby_question">
<owl:DatatypeProperty rdf:about="questiontext_question">
<rdfs:range rdf:resource="http://www.w3.org/2001/XMLSchema#string"/>
<owl:Restriction>
<owl:minCardinality
<owl:onProperty rdf:resource="questiontext_question"/>
</owl:Restriction>
<owl:DatatypeProperty rdf:about="usecase_question_shortanswer">
<rdfs:range rdf:resource="http://www.w3.org/2001/XMLSchema#short"/>
<owl:Restriction>
<owl:minCardinality
<owl:onProperty rdf:resource="usecase_question_shortanswer"/>
</owl:Restriction>
<owl:DatatypeProperty rdf:about="answer_question_answers">

166

<owl:Restriction>
<owl:minCardinality
<owl:onProperty rdf:resource="answer_question_answers"/>
</owl:Restriction>
<owl:DatatypeProperty rdf:about="tolerance_question_numerical">
<owl:Restriction>
<owl:minCardinality
<owl:onProperty rdf:resource="tolerance_question_numerical"/>
</owl:Restriction>
<owl:DatatypeProperty rdf:about="single_question_multichoice">
<rdfs:range rdf:resource="http://www.w3.org/2001/XMLSchema#short"/>
<owl:Restriction>
<owl:minCardinality
<owl:onProperty rdf:resource="single_question_multichoice"/>
</owl:Restriction>

Figure 9. The OWL code of a sub-part of the Question's module in the Moodle Database

5. CONCLUSION AND FURTHER WORKS
In this paper, we have proposed a novel method for generating an e-Learning ontology using a
technique which consists in a set of transformation rules for building OWL ontologies from
relational databases. The schema mapping uses the transformation rules to transform the
components of the physical model into ontology's components. The data analysis is used to
recover some disappeared aspects during mapping conceptual data model to the relational model
(like disjointness and totalness in simple inheritance cases and the participating level of tables in
n-ary relations).
We have implemented our proposal using the Moodle database of the pedagogical platform of the
CA University. The obtained results are satisfactory compared to other methods in terms of the
number of the treated concepts. Moreover, our method covers all possible cases in databases. The
generated ontology has richer non-taxonomic relations.
A major direction for improvement could be to add a reverse engineering phase before applying
the transformation rules in order to detect generalization and specialization inheritance cases. This
process will allow us to recover disappeared tables during mapping conceptual data model to the
relational model. Furthermore, integrating a reverse engineering phase will make the built
ontologies richer in terms of taxonomic relations. Finally, we also suggest refine the obtained
ontology by renaming its components in order to make them more expressive.

ACKNOWLEDGMENTS
We want to thank Misters Driss Goujdami and Abdelali Rochdi for providing us a dump file of
the database of the pedagogical platform of the Cadi Ayyad University. We also want to thank
Mister My. Mehdi Bouhamidi for his technical support to deploy the dump file in another
database server.

167


REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]

W3C, the semantic web activity, Available at : http://www.w3.org/2001/sw/
T. Gruber, Ontology, In Ling Liu & M. Tamer Özsu (Ed.), the Encyclopedia of Database Systems
(Springer-Verlag, 2009, 1963-1965).
Drucker, P. (2000). Need to Know: Integrating e- Learning with High Velocity Value Chains. A
Delphi Group White Paper. Available at : http://www.delphigroup.com/whitepapers/pdf/20001213-elearning-wp.pdf
Stephen Downes, E-learning 2.0 eLearn Lagazine, October 2005. Available at:
http://elearnmag.acm.org/featured.cfm?aid=1104968
H. Moutachaouik, H. Douzi, A. Marzak, H. Behja, B. Ouhbi, (2012) “ Plugin of Recommendation
based on a hybrid method for the ranking of documents in the e-learning platforms”, 5th International
Conference, ICISP 2012, pp 587-595.
M. Pazzani & D. Billsus. (2007) The Adaptive Web, chapter Content-Based Recommendation
Systems, pages 325–341. Springer Berlin / Heidelberg.
R. Burke, K. Hammond, and E. Cooper. (1996) “Knowledge based navigation of complex
information spaces”, In Proc. of the 13th National Conference on Artificial Intelligence (AAAI’96),
pages 462–468.
D. Goldberg, D. Nichols, B.M. Oki, and D. Terry. (1992) “Using Collaborative filtering to weave an
information tapestry”, Communications of the ACM, Vol. 35, No. 12, pp. 61–70.
Khribi, M. K., Jemni, M., & Nasraoui, O. (2009). “Automatic Recommendations for E-Learning
Personalization Based on Web Usage Mining Techniques and Information Retrieval”, Educational
Technology & Society, Vol. 12, No. 4, pp. 30–42.
Brut, M.; Sèdes, F., (2010) “Ontology-Based Solution for Personalized Recommendations in ELearning Systems. Methodological Aspects and Evaluation Criterias”, Advanced Learning
Technologies (ICALT), 2010 IEEE 10th International Conference on, pp.469-471.
Huiyi Tan; Junfei Guo; Yong Li, (2008) “E-learning Recommendation System”, Computer Science
and Software Engineering, 2008 International Conference on , vol.5, pp.430,433.
Alexander Maedche & Steffen Staab, (2005) “Ontology Learning for The Semantic Web”, IEEE
Intelligent Systems, Vol. 16, No. 2, pp. 72-79.
A. Gomez-Perez, D. Manzano-Macho, A Survey of Ontology Learning Methods and Techniques,
Deliverable,
1.5.
OntoWeb
Consortium,
(2003).
Available
from:
http://www.stiinnsbruck.at/fileadmin/documents/deliverables/Ontoweb/D1.5.pdf, last visited: 03.15.2010
R. Chiang, T. Barron, V. Storey, Reverse engineering of relational databases: Extraction of an EER
model from a relational database, Data & Knowledge Engineering, Vol. 12, n. 2, pp. 107–142, 1994
U. Fayyad, G. Piatetsky-Shapiro, P. Smyth, From Data Mining to Knowledge Discovery in
Databases, AI Magazine, Vol. 17, pp. 37-54, 1996
Chen He-ping, He Lu & Chen Bin, (2008) “Research and implementation of ontology automatic
construction based on relational database”, Proceedings of the International Conference on Computer
Science and Software Engineering, pp. 1078-1081.
Heru Agus, Su-Cheng Haw & Ziyad.T. Abdul-Mehdi, (2011) “Ontology extraction from relational
database: Concept hierarchy as background knowledge”, Knowledge-Based Systems, Vol. 24, No 3,
pp. 457-464,
Justas Trinkunas & Olegas Vasilecas, (2007) “Building ontologies from relational databases using
reverse engineering methods”, Proceedings of the international conference on Computer systems and
technologies.
Farid Cerbah, (2008) “Mining the Content of Relational Databases to Learn Ontologies with Deeper
Taxonomies”, Proceedings of the 2008 IEEE/WIC/ACM International Conference on Web
Intelligence and Intelligent Agent Technology, pp. 553- 557.
Zhuoming Xu, Xiao Cao, Yisheng Dong & Wenping Su, (2004) “Formal Approach and Automated
Tool for Translating ER Schemata into OWL Ontologies”, Advances in Knowledge Discovery and
Data Mining, pp. 464-475.
Igor Myroshnichenko & Marguerite C. Murphy, (2009) “Mapping ER Schemas to OWL Ontologies”,
Proceedings of the IEEE International Conference on Semantic Computing, pp. 324-329.
Sujatha R. Upadhyaya & P. Sreenivasa Kumar, (2005) “ERONTO: A Tool for Extracting Ontologies
from Extended ER Diagrams”, Proceedings of the 20th ACM Symposium on Applied Computing, pp.
666-670.
168


[23] Shihan Yang & Jinzhao Wu, (2010) “Mapping Relational Databases into Ontologies through a Graphbased Formal Model”, Proceedings of the Sixth International Conference on Semantics Knowledge
and Grid, pp. 219-226.
[24] Christian Bizer, (2003) “D2R MAP - a database to RDF mapping language”, Proceedings of the 12th
International World Wide Web Conference.
[25] Nikolaos Konstantinou , Dimitrios-emmanuel Spanos , Michael Chalas , Emmanuel Solidakis &
Nikolas Mitrou, (2006) “VisAVis: An Approach to an Intermediate Layer between Ontologies and
Relational Database Contents”, Proceedings of Workshops and Doctoral Consortium, The 18th
International Conference on Advanced Information Systems Engineering - Trusted Information
Systems.
[26] Shihan Yang, Ying Zheng & Xuehui Yang, (2010) “Semi-automatically building ontologies from
relational databases”, Proceedings of the 3rd IEEE International Conference on Computer Science
and Information Technology, pp. 150 - 154.
[27] Jesús Barrasa , Óscar Corcho & Asunción Gómez-pérez, (2004) “R2O, an Extensible and
Semantically based Database-to-Ontology Mapping Language”, Proceedings of the 2nd Workshop on
Semantic Web and Databases, pp. 1069-1070.
[28] Irina Astrova , Nahum Korda & Ahto Kalja, (2007) “Rule-Based Transformation of SQL Relational
Databases to OWL Ontologies”, Proceedings of the 2nd International Conference on Metadata &
Semantics Research, pp. 415-424.
[29] Man Li, Xiao-Yong Du & Shan Wang, (2005) “Learning ontology from relational database”,
Proceedings of the International Conference on Machine Learning and Cybernetics, pp. 3410-3415.
[30] Zdenka Telnarova, (2010) “Relational database as a source of ontology creation”, Proceedings of the
International Multiconference on Computer Science and Information Technology, pp. 135-139.
[31] Kobra etminani , Mohsen Kahani & Yanehsari N.R, (2009) “Building ontologies from relational
databases”, Proceedings of the First International Conference on Networked Digital Technologies, pp.
555-557.
[32] Changjun Hu, Huayu Li, Xiaoming Zhang & Chongchong Zhao, (2008) “Research and
Implementation of Domain-Specific Ontology Building from Relational Database”, Proceedings of
the Third ChinaGrid Annual Conference, pp. 289 - 293.
[33] Ahmed Waqas, Aslam Muhammad Ahtisham, Shen Jun & Yong Jianming, (2011) “A light weight
approach for ontology generation and change synchronization between ontologies and source
relational databases”, Proceedings of the 15th International Conference on Computer Supported
Cooperative Work in Design, pp. 208-214.
[34] Nadine Cullot, Raji Ghawi & Kokou Yétongnon, (2007) “DB2OWL: A Tool for Automatic Databaseto-Ontology Mapping”, Proceedings of the 15th Italian Symposium on Advanced Database Systems,
pp. 491-494.
[35] Irina Astrova & Ahto Kalja, (2006) “Towards the Semantic Web: Extracting OWL Ontologies from
SQL Relational Schemata”, Proceedings of the 2nd International Conference on Metadata &
Semantics Research, pp. 62-66.
[36] Guohua Shen, Zhiqiu Huang, Xiaodong Zhu & Xiaofei Zhao, (2006) “Research on the Rules of
Mapping from Relational Model to OWL”, Proceedings of the OWLED*06 Workshop on OWL:
Experiences and Directions.
[37] Ramez Elmasri & Shamkant Navathe, (2011) Fundamentals of Database Systems sixth Edition
Pearson Education.

169

A novel method for generating an elearning ontology

Recommended

More Related Content

What's hot (19)

Viewers also liked (9)

Similar to A novel method for generating an elearning ontology (20)

Recently uploaded (20)

A novel method for generating an elearning ontology