[ieee 2012 12th international conference on computational science and its applications (iccsa) -...
TRANSCRIPT
Pedagogical Model based on Semantic Web Rule Language
Camila Bezerra da Silva
Center for Exact Sciences and TechnologyFederal University of Recncavo of Bahia - UFRB
Cruz das Almas - [email protected]
Abstract—Intelligent Tutoring System(ITS) aims to providepersonalized instruction. Its main module, pedagogical model,is responsible for choosing which strategy to apply to aspecific student, as well as determining which learning topicsto provide. However, there are several strategies and teachingtechniques, and each student has his peculiarities. This paperproposes the use of Semantic Web Rule Language(SWRL) toassist ITS in the making decision on which learning strategyto apply to a specific student in a given time. SWRL isbased on Horn-like rules expressed in terms of Ontology WebLanguage(OWL) concepts.
Keywords-Learning; Ontology; SWRL; Tutoring;
I. INTRODUCTION
Concerning to Intelligent Tutoring Systems(ITS), the pe-
dagogical model is responsible for choosing the strategy and
tactics that will be applied to a particular student, based on
the interaction that this student will have with the system.
The main goal of ITS is to provide a personalized teaching to
the student. Therefore, it is necessary to know the student’s
needs, and through the student profile to infer what strategy
to apply in a certain time.
In last decades, pedagogy researchers have developed
many teaching strategies, aiming to make learning activities
more efficient. Some of these strategies have been persistent
over the years, as the strategies had been based on the
behaviorist and constructivist theories. Computer systems
applied to education have used these theories. However,
some questions are still open: Which strategy to apply to
a student in a given time and when to change the strategy?
Which topics on the subject to be studied display? In what
order?
In order to answer these questions, it has been used
Artificial Intelligence techniques such as ontologies, rules-
based systems and intelligent agents, and with these tech-
niques such good results have been achieved( [1], [2]).
Ontologies are used to represent knowledge of a domain. In
the case of tutoring systems, ontologies attend to represent
the knowledge to be taught and the learner profile.
Following this line, this paper proposes the use of on-
tologies and Semantic Web Rule Language(SWRL) to build
the pedagogical model. SWRL emerged from deficiency of
ontologies that can not represent Horn-like rules. SWRL
rules are written in the antecedent-consequent form. The
antecedent and consequent consist of a conjunction of one
or more atoms expressed in terms of OWL concepts.
Thus, by using a rule engine like Jess [3], it is possible
to reason about such rules and answers the questions posed
previously, i.e, to know what strategy to apply to a student
at any given time, according to his interaction.
This paper is organized as follows: section II describes
some of main concepts about Intelligent Tutoring Systems
and Pedagogical Theories; section III presents Ontologies
and the Semantic Web Rule Language(SWRL); section IV
describes the pedagogical model based on SWRL; section V
is devoted to related work, and section VI finish this paper
with some conclusions.
II. INTELLIGENT TUTORING SYSTEMS
The Intelligent Tutoring Systems(ITS) appeared in the
90’s with the main objective of providing tools that meet
the individual needs of each student.
ITS consists of computer systems that support learning
and try to adapt to the needs of the particular student.
These systems incorporate Artificial Intelligence techniques
for knowledge representation, reasoning and user interaction.
Moreover, ITS offers an individualized learning process
according to student’s needs. To achieve this goal, ITS
knowledge base is updated and adapts its teaching strategies
according to student interaction.
According to [4], [5] and [6] the key features of an ITS
are:
• The domain knowledge is represented explicitly and
unambiguously, being possible to make inferences on
such knowledge.
• The student acquires knowledge that allows the ITS to
manage and adapt learning.
• The sequence of teaching is not predetermined, it can
adapt.
• It can generate appropriate problems, tips and it adapts
to student’s needs.
These systems have basic architecture consists of four
components: student model, pedagogical model, domain
model and interface, as shown in Figure 1.
2012 12th International Conference on Computational Science and Its Applications
978-0-7695-4710-7/12 $26.00 © 2012 IEEE
DOI 10.1109/ICCSA.2012.31
125
Figure 1. architecture
In the student model is represented the knowledge, behav-
ior and cognitive characteristics of the student. The domain
model contains the domain knowledge of the subjects to
be taught or learned. The pedagogical model contains the
teaching strategies and tactics that are used according to
student interaction with the system. Finally, the interface
mediates the interaction between tutor and a student.
This paper is centered in the pedagogical model that is
responsible for the organization and selection of teaching
strategies and tactics to be used in interaction with the
student, thus providing a personalized teaching.
In the next section, it was described the main pedagogical
theories.
A. Pedagogical Theories
There is still no pedagogical theory for distance learning.
However, it has been adopted traditional pedagogical the-
ories, which are already consolidated, and they have been
applied in the teaching process in general. The trend is the
use flexible of different pedagogical theories, to meet the
diversity of resources and students.
The most popular theories are Behaviorist, Constructivist-
Interactionism and Constructivist Social-Interactionism, and
in each of these theories, their representatives stand out:
Skinner, Piaget and Vygotsky, respectively.
The Behavourism [7] had its origin in the 50’s, when pro-
fessor Skinner of Harvard has proposed a teaching machine.
According to this theory, students are taught in way that they
are induced to engage in new forms of behavior and specific
ways. The material to be taught is transmitted in a linear
and sequential way. By allowing students to interact with
a learning content, it is not stimulated learner autonomy,
which is faced with a sequential structure.
With the crisis of the behaviorist paradigm, came into play
in the mid 50’s: the cognitive psychology, which the main
feature is the construction of knowledge through information
processing. From this, emerged a new approach to teaching:
the Constructivism.
The Constructivist-Interactionism theory suggests that the
learner understands the world through his perceptions, by
building meant for this world [7].
Piaget, its main defender, believed that learning occurs
in stages that are directly linked to the mental development
of each student. This theory are centered on individual de-
veloping, each learner should construct his own knowledge,
without take account the social-history context.
In this context, the goal of the teacher would be to
encourage individual discovery, and no more than determine
the speed and way of construction of knowledge of the
student.
Parallel to the development of Constructivist-
Interactionism, computer systems have emerged with
Artificial Intelligence, and appeared several studies that
relate Education, Web and Artificial Intelligence. Such area
that came up with this joint is called e-learning.
Finally, the Social-Interactionism approach proposed by
Vygotsky had as the main objective the interaction between
an individual with others [7]. According to this theory,
human intelligence is constituted by cultural tools, such
as language that are the legacy of past generations and,
therefore, can only be understood from a socio-historical
perspective of human cognition.
Vygotsky proposed the concept of Zone of Proximal
Development, where he said that what an individual ac-
complishs with the help of another individual, as a tutor
or a friend, or even resources such as books, videos and
computer, it also represents an intellectual ability of the
individual. Unlike the Piaget’s approach which consider as
intellectual ability, only what each individual was able to
build by himself without interactions with others.
The role of the teacher in this context is to promote
social harmony by stimulating the exchange of information
in search of building shared knowledge. Furthermore, the
computation is now seen as a means of communication
between tutors and learners.
Besides these theories, there are several tactics of teaching
that can be used in the learning process. Each tactic is related
to one or more theories. So for every theory, there are several
tactics that can be employed in learning.
The problem is to know which tactics to apply to a
student, by considering that each student has own skills and
deficiencies.
To solve this problem, this paper proposes to use Ontolo-
gies and Semantic Web Rule Language, which it is explained
in the next section.
126
III. ONTOLOGY AND SEMANTIC WEB RULE LANGUAGE
An ontology can be described as an explicit and formal
specification of a domain which someone has an interest
in representing. Ontologies are based on Description Logic
that consist of concepts and relationships between these
concepts that are organized in a hierarchical structure [8].
Moreover, it is worth highlighting that it is possible to
represent disjunction and combination of classes, property
restrictions and characteristics of such properties.
The set of concepts and their relationships of domain
are called knowledge base. Through this knowledge base, a
computer system can interpret it and derive new knowledge,
by returning useful information as well as inferring implicit
information to the user.
Usually, in a certain area of knowledge, different experts
have different understanding of the concepts involved, which
leads to problems will be in communication, and ambiguity.
Ontologies provide standardization to e-Learning systems,
by reaching a shared vocabulary consensus in their unders-
tanding of a domain.
OWL is the language recommended by The World Wide
Web Consortium (W3C) [9] to the specification of ontologies
for the emerging Semantic Web technologies [10]. OWL is
a family of knowledge representation with different levels
of expressivity: OWL-Lite, OWL-DL and OWL-Full. This
paper focuses on OWL-DL which is based on Description
Logics(DL) [11]. Besides, OWL-DL has a well-defined
syntax, a formal semantic, convenience of expression and
support for reasoning efficiently.
An OWL-DL ontology, as a DL based ontology, does
not support rules, which limits expressivity. There are cases
where it is need to express some definitions and this is only
possible with such rules.
Consider an ontology for the domain of a family. In
the Description Logic used to specify ontologies, it is not
possible to have a rule that expresses “the brother of the
father of a person is his uncle”. To accomplish this, it came
up the Semantic Web Rule Language(SWRL) that is also
recommended by W3C.
SWRL [12] is a rule description language which
combines RuleML and OWL. It allows users to create rules
in the Horn clauses format [13] that can be expressed in
terms of OWL classes. In addition, it can reason about
OWL individuals. Thus, with rule like below, it can be
solved the problem described in the example above:
father(?x, ?y) ∧ brother(?x, ?z) ∧ man(?z) →uncle(?z, ?y)
What means that a person ?y that has a father ?x and
this father ?x has a brother ?z, that implies the person ?z is
his uncle.
SWRL rules are in the form antecedent → consequent,
which is interpreted so that if the antecedent is true, then
the consequent must also true. Moreover, both the antecedent
and the consequent parts are conjunctions of atoms, accor-
ding the Horn clauses [13].
In this context, ontologies are used in e-Learning to
represent the learning domain, and SWRL are used to
express rules that can not be expressed in OWL language.
In the next section, it will be described the pedagogical
model based on semantic web rule language.
IV. PEDAGOGICAL MODEL BASED ON SEMANTIC WEB
RULE LANGUAGE
As we saw, there are some consolidated learning theories.
Each one with its own characteristics, and provide the
student a different way of learning. However, each student
has a different way of assimilating the content to be studied,
some learning more by studying alone, some in groups, some
reading, some watching a presentation, etc...
The questions that arise are: What strategy to apply to
each student? How can offer personalized learning? What
sequence of learning topics offer?
These are issues that this paper proposes solve, by using
Semantic Web technologies.
The first step is to identify characteristics of the student,
cognitive aspects and about the domain to be studied. This
process results in the student model, which is represented
by ontologies. The student model is updated according to
student interaction with the system.
For this, a preliminary analysis is performed, which it is
given a questionnaire to the student with questions about the
learning domain and cognitive aspects.
Both the questionnaire follows the IMS as the ontology
which represent the student model [14]. The IMS specifi-
cation defines several profiles of information that must be
obtained from the student in order to take a better knowledge
of the capabilities, intentions and actions of this.
In addition, it is being used questions of the Learning
Styles Questionnaire( [15]) that defines several of questions
to infer the type of student.
The original questionnaire is in portuguese language, and
adopted has questions such as:
What content topics the student has studied?
What the student has not studied?
Do you prefer study alone or in a group?
Do you prefer study with a linear content module or a non-
linear content like hypertexts?
Do you learn better texts with pictures and diagrams inside
or not?
Based on the answers of the questions, an initial profile
of the student is created, and selected which strategy and
tactics to use, initially. Therefore, there is a relationship
between strategies and tactics that are represented in the
127
Figure 2. Pedagogical Model
ontology, moreover questions and their alternatives are also
represented in this ontology.
The strategies considered in this work are Be-
haviorist, Constructivist-Interactionism and Constructivist
Interactionism-Social, that were explained in the previous
section II-A. There are several teaching tactics that are
consistent with these strategies, and so, these tactics are
grouped by strategy.
The ontology was built using the Protege [16], which is
an ontology editor. Part of the ontology that exhibits the
relationship between the strategies and tactics, is showing
in Figure 2.
The questions are presented to the student in a non-
linear sequence, the initial questions related to student
cognition, aims to find out which strategy fits more with
the student. For example, by considering the question below:
I prefer studya).a linear learning module, in other words, i like thatlearning content be given to me in a sequential andpredefined structure.b).a free content, choosing the topics by self.
If the student chose the first option, it is possible to
infer that he fits better with behavioral strategy, which
believes that students should follow a sequential structure,
linear and predefined. Otherwise, if he chooses the second
option, certainly the most appropriate strategy would be
constructivist.
It was noted that such decisions about which strategy
and/or tactics would be best interpreted using SWRL, which
as already stated, is a language for building rules like:
premises → conclusion, following the definition of the Horn
clauses.
So, the question above can be expressed in SWRL:
Figure 3. SWRL Rules
Person(?x) ∧ Question(I prefer to study) ∧Choice(a linear learning module) →applyStrategy(Behaviorist, ?x)
Which means that IF a person X AND the question is “I
prefer to study” AND the choice was “a linear learning
module” THEN apply the Behaviourist strategy for person
X.
Inside the Protege, there is the “SWRL Tab” that is
an extension to the Protege-OWL plugin that permits the
creation and execution of SWRL rules. Some rules of
pedagogical model are shown in Figure 3.
To run this rules, and so infer what strategy and tactics to
apply, it was used the SWRLJessTab that is a plugin to the
SWRLTab in Protege that supports the execution of SWRL
rules using the Jess rule engine [3].
After create the OWL concepts and necessary SWRL rules
in the Jess, it can be perform inference. When execute the
rules, new Jess facts are inserted into the fact base, and those
facts can be used in further inference.
A. Preliminary Results
This proposed application has been applied to Brazilian
students of course of exact sciences and technology, of the
Federal University of Reconcavo of Bahia.
In the present stage, it was carried out a part of the tests,
which it was analyzed the performance of the application
with 30 students of the mentioned course. The students are
between 17 and 35 years, and most are male.
The pedagogical model was applied to them, and also
the students were asked about the tactics of teaching that
they learned better. Then, it was verified if the tactics
recommended by the model, in fact, matched the tactics
outlined by the student.
In these previous tests, the proposed model achieved an
average of 85% accuracy.
V. RELATED WORK
During the last decades, there are an increasing and active
researches about computer systems to Education. And a lot
of events was held about this theme.
Recently, [17] propose a system and a set of strategies that
provide a way to generate, automatically, multiple choice
questions from SWRL rules. It has been designed and
implemented as a framework that support new question and
answer formation strategies. The system consists of three
layers. The first layer is responsible for parsing rule files and
to convert them into an internal representation. The second
128
layer is responsible for generating a textual representation
for each variable. The last layer is the one responsible for
the generation of the appropriate distracting answers. The
system is implemented in such a way that it possible to
addition new strategies.
Since [18], defends the idea of using SWRL rules to
represent teaching strategies and convert them to Jess for
the environment where they can be tested. The conversion
mechanism makes use of existing tools, like SweetRules that
provide a suite of converters among several rule languages
and rule engines.
Despite the importance of these works to area, none
of them, it was taken into account the different teaching
strategies, much less been a survey of tactics that can be
used in conjunction with such strategies. Moreover, it was
not analyzed the characteristics of each student, to resulting
in the student model, and thus by helping to provide a
personalized learning.
VI. CONCLUSION AND FUTURE WORK
As shown in this paper, Web Semantic technologies can
be applied for Education. The main module of Tutoring
Intelligent System is the pedagogical model, which is re-
sponsible for choosing the strategy and tactics that will be
applied to a particular student, according to his interaction
with the system. The most popular learning theories are
Behaviorist, Constructivist-Interactionism and Constructivist
Social-Interactionism.
As we seen, this can be done with Semantic Web Rule
Language(SWRL) that is a language to write rules-like Horn.
To run this rules, and so infer what strategy and tactics to
apply, it was used a rule engine, the Jess, where users can
run SWRL rules interactively to create new OWL concepts
and then insert them into an OWL knowledge base, or just
verify the validity of the rule.
Thereat, it was concluded that Web Semantic technologies
can be applied efficiently to meet the goals of Tutoring In-
telligent Systems, in particular, to build pedagogical model.
Regarding to future work, we intend to improve the form
of student assessment, by building a more robust model
of the student. Thus, the pedagogical model may suggest
actions with more certainty.
REFERENCES
[1] N. Henze, P. Dolog, and W. Nejdl, “Reasoning and ontologiesfor personalized e-learning in the semantic web,” EducationalTechnology & Society, vol. 7, pp. 82–97, 2004.
[2] L. M. M. Giraffa and R. M. Vicari, “The use of agentstechniques on intelligent tutoring systems,” in SCCC, 1998,pp. 76–83.
[3] M. OConnor, H. Knublauch, S. Tu, and M. M. Musen,“Writing rules for the semantic web using swrl and jess,”in Protg with Rules Workshop, held with 8th InternationalProtg Conference, Madri, 2005.
[4] M. Urretavizcaya, “Presentacion monografia: Sistemas in-teligentes en el ambito de la educacion,” Inteligencia Artifi-cial, Revista Iberoamericana de Inteligencia Artificial, vol. 5,no. 12, pp. 2–4, 2001.
[5] S. Jonassen, D.H. & Wang, “The physics tutor: Integratinghypertext and expert systems,” Journal of Educational Tech-nology Systems, vol. 22, pp. 19–28, 1993.
[6] I. Arroyo, K. Ferguson, J. Johns, T. Dragon, H. Mehera-nian, D. Fisher, A. Barto, S. Mahadevan, and B. P. Woolf,“Repairing disengagement with non-invasive interventions,”in 13th International Conference of Artificial Intelligence inEducation., 2007.
[7] J. I. Pozo, Teorias Cognitivas da aprendizagem, 3rd ed. PortoAlegre: Artmed, 1998.
[8] G. Antoniou and F. van Harmelen, A Semantic Web Primer,J. W. Michael Papazoglou and J. Mylopoulos, Eds. MITPress, 2004.
[9] S. Bechhofer, F. van Harmelen, J. Hendler, I. Horrocks, D. L.McGuinness, P. F. Patel-Schneider, and L. A. Stein, “OWLWeb Ontology Language reference,” W3C Recommendation,10 February 2004, available at http://www.w3.org/TR/owl-ref/. [Online]. Available: http://www.w3.org/TR/owl-ref/
[10] P. Hitzler, M. Krotzsch, and S. Rudolph, Foundations ofSemantic Web Technologies. Chapman & Hall/CRC, 2009.
[11] F. Baader, I. Horrocks, and U. Sattler, “Description Logics,”in Handbook of Knowledge Representation, F. van Harmelen,V. Lifschitz, and B. Porter, Eds. Elsevier, 2008, ch. 3, pp.135–180. [Online]. Available: download/2007/BaHS07a.pdf
[12] I. Horrocks, P. F. Patel-Schneider, H. Boley, S. Tabet,B. Grosof, and M. Dean, “SWRL: A semantic webrule language combining OWL and RuleML,” W3CMember Submission, 21 May 2004, available at http://www.w3.org/Submission/SWRL/. [Online]. Available: http://www.w3.org/Submission/SWRL/
[13] C.-L. Chang and R. C.-T. Lee, Symbolic Logic and Mechani-cal Theorem Proving, 1st ed. Orlando, FL, USA: AcademicPress, Inc., 1997.
[14] I. G. L. Consortium, “Ims learner information package speci-fication,” http://www.imsglobal.org/profiles/index.html, 2005,accessed in january 2011.
[15] R. M. Felder and R. Brent, “Understanding student differ-ences,” Journal of Engineering Education, vol. 94, pp. 57–72,2005.
[16] Stanford, “Protege ontology editor and knowledge acquisi-tion system,” http://protege.stanford.edu, Stanford University,2011, accessed in january 2011.
[17] K. Zoumpatianos, A. Papasalouros, and K. Kotis, “Automatedtransformation of swrl rules into multiple-choice questions,”in FLAIRS Conference, 2011.
[18] E. Wang and Y. S. Kim, “A teaching strategies engine usingtranslation from swrl to jess,” in Intelligent Tutoring Systems,2006, pp. 51–60.
129