A WEB-BASED MULTILINGUAL INTELLIGENT TUTOR
SYSTEM BASED ON JACKSON’S LEARNING STYLES
PROFILER AND EXPERT SYSTEMS
H. MOVAFEGH GHADIRLI
Young Researchers Club, Islamshahr Branch, Islamic Azad University, Islamshahr, Iran
[email protected]
M. RASTGARPOUR
Department of Computer Engineering, Saveh branch, Islamic Azad University, Saveh,
Iran
[email protected]
Nowadays, Intelligent Tutoring Systems (ITSs) are so regarded in order to improve
education quality via new technologies in this area. One of the problems is that the
language of ITSs is different from the learner’s. It forces the learners to learn the system
language. This paper tries to remove this necessity by using an Automatic Translator
Component in system structure like Google Translate API. This system carry out a pre-
test and post-test by using Expert System and Jackson Model before and after of training
a concept. It constantly updates learner model to save all changes in learning process. So
this paper offers an E-Learning system which is web-based, intelligent, adaptive,
multilingual and remotely accessible where tutors and learners can have non-identical
language. It is also applicable Every Time and Every Where (ETEW). Furthermore, it
trains the concepts in the best method with any language and low cost.
1. Introduction
In 1982, Sleeman and Brown reviewed the state of the art in computer-aided
instruction. They proposed a novel term of ITSs to describe these evolving
systems as well as distinguish them from other educational systems. So the
implicit assumption corresponding to the learner focused on learning-by-doing.
They [1] classified available ITSs into four categories include of problem-
solving monitors, coaches and teachers, laboratory instructors, and consultants.
The first E-Learning software was static and non-intelligent. A course had been
organized by a prior procedure and taught in the same style for all learners. They
were either computerized versions of textbooks (characterized as electronic page
turners) or drill and practice monitors which were giving a learner some
problems and comparing the learner’s answers with pre-scored answers[1].
1
2
There are several kinds of Learners in the Web. It means that some learners
need to repeat some lessons mean while some lessons must be removed for
others. Later the researchers concluded that learning process must be dynamic
and intelligent based on pedagogy view [2]. But developing an applicable and
trustful system is so hard [2]. This caused to advent new generation of intelligent
educational systems [3].
Web-based learning and intelligent learning is so regarded in education
todays[2, 4]. A web-based tutor has some benefits like tirelessly, dominance on
concepts, low cost and independent of time and place. It can utilizes
conversational agents as well [3]. However millions learners of the world can
learn via thousands of expert tutor through the web in an intelligent and virtual
school. One of the important classes of intelligent tutors is based on rule-based
expert system. It determines whatever the student knows, doesn’t know and
knows incorrectly. It uses this information to adjust learning style. This
education style which enjoys the benefits of Expert Systems is called model
tracing tutor [5].
ITS sometimes has different language of learner language. So such
participate needs to know its language in this virtual class. On the other hand, the
learning of a foreign language is often difficult and impractical for adults.
Moreover, many such courses can be very time consuming, because learners
often must cope with unfamiliar writing systems as well as differing cultural
norms. So a system is demanded in order to eliminate this “language barrier”.
This paper introduces an intelligent system to enjoy Expert Systems
abilities. So E-Learning would be efficient, adaptive and just needs a computer
and internet connection. Adapting with web-based contexts is very important,
because the concept, which is developed for one user, isn’t useful for others [2].
The proposed system determines the learning style through a test. It
develops a primary model of learner. The learning process starts then. Some
characteristics of learner may be varying during learning progress gradually.
These progresses would be saved in learner model by system. So learner model
gets closer step by step. The system can receives scientific and mental feedback
of the learner and change the learning style during the process.
A learner doesn’t have to know extra language by using an Automatic
Translation component. The learner can be read the contents of E-books, virtual
blackboard, chat rooms and even write with any known language.
Web-based learning also is useful for training new employees and adapting
staffs according to company changes. However proposed system is an artificial
environment that easily loses the motivational benefits of authentic task-oriented
dialog, but ideally, web-based learning improves the learning efficiency for
3
students and employees through features that are not available in face to face
learning. It also allows learners to access the learning materials and interact with
the rest of course ETEW. The aim of proposed system is to offer the content
which the learner can’t be aware of it easily with any language.
The rest of this paper is organized as follows. Section II defines an
intelligent tutoring system, automatic translator and presents some available
samples. It also deliberates the role of pedagogy in learner modeling Jackson
Model. Section III describes the proposed E-Learning system which is
intelligent, adaptive, customized, web-based and multilingual. Finally this paper
concludes in Section IV.
2. Background
2.1. Intelligent Tutoring System
ITSs are computer-based instructional systems with models of educational
content. They specify what to teach, and also teaching strategies that specify how
to teach [6]. In late 1960, the ITSs have moved out of academic labs and have
been applied in classrooms and workplaces. Some of them have shown to be
high effective [5]. Unfortunately, intelligent tutors are difficult and expensive to
build whereas they are more common and have proven that to be more effective.
Intelligent systems can recognize the learner type, choose appropriate course
content from knowledge base and present it to learners in proper style. It also
attempts to simulate a human tutor expertly and intelligently. Students using
these systems usually solve problems and related sub-problems within a goal
space, and receive feedback when their behavior diverges from that of the
learner model.
Some design factors have role for each ITS which including component
expert simulator, tutor software, learner model, modeler, and knowledge base.
Error! Reference source not found. illustrates them.
ITSs allow “mixed-initiative” tutorial interactions, where learners can ask
questions and have more control over their learning. Some available ITSs are
introduced in Error! Reference source not found.1.
Some web-based ITSs are presented in Table 2.
2.2. Automatic Translation
Automatic Translating System is more efficient than professional human
translators in terms of cost and speed of translation. Two types of translation
tools which are used frequently are as follows:
4
• Machine Translation is suitable for e-learning localization and multilingual
information retrieval.
• Translation Memories takes a certain source text and then stores it with
corresponding translation done by a human. It can automate the translation
process completely in desirable mode.
Machine translation has been appeared in 1981 for personal computers. In 1997,
Babel Fish appeared the first, free and translation service on the World Wide
Web [7].
Figure 1. Intelligent Tutoring System Components
The Automatic Translator includes a Natural Language Parser (NLP). It can
annotate phrases with structural information and refer to relevant grammatical
explanations. An Error Model which detects and analyzes syntactic mistakes
would be referred as well.
Table 1. Some Available Intelligent Tutoring Systems [8]
Name Comments
CAES system [9] Has been developed by integration of shipping simulation and
intelligent decision system.
ICATS [10] Coordinates an expert system with multimedia system in an intelligent
learning system.
ANDES [11] Trains Physics without using natural language.
ATLAS [12] Uses natural language to teach Physics. It can answer to learner’s
questions.
ELM-ART [13] A web-based tutoring System for learning programming in LISP.
APHID-2 [14] A hypermedia generation system with adaption by defining rules that
map learner information.
AUTOTUTOR [15] Be full intelligent and teach “introduction to computers” in at least 200
universities of the world.
5
The main task of the Automatic Translator component is real-time
translating of text contents from one language to another. The teacher can store
contents with any language in the system without user intervention. It will
translate to the learner language then.
Table 2. Adaptive and Intelligent Technologies in Web-based ITS SYSTEMS [16]
Presentation
Adaptive
Component
Sequencing
Intelligent
Hypertext
Adaptive
Analysis
Problem
solution
Support
Solving
System
AST Y Y Some
InterBook Y Y Server Y Some
PAT-InterBook Y Y
DCG Y
PAT N Y
WITS Y Y
C-Book Y Some
Manic Y Some Server Y Y
Proposed System Y Y Some
It should be noted that multilingual meetings usually is online and also
involve more than two learners with different languages. Automatic translation
can be applicable in chat among the learners. Table 3 presents some applications
which provide automatic translation for instant messaging (Chat).
Table 3. Online Chat Systems with Automatic Translation [17]
Application Languages
Amikai 9
Annochat 4
ChatTranslator 7
Free2IM Hab.la 13
Realtime Chat 41
IBM Lotus 7
Sametime 7
MeGlobe 15
WorldLingo Chat 15
Any Automatic Translator needs to a web-based multilingual dictionary to
be useful for online and real-time multilingual translation. Nowadays, the most
6
of web services are provided by online dictionaries supporting two or more
languages.
One of the most popular web based dictionary is Google Translate to
translate each pair of 64 languages to each other such as English to Spanish,
French to Russian, Chinese to English, etc. There are 4032 language pairs to
translate.
Google Translate* is based upon a statistical translation system unlike other
available machine translation systems. In fact a language model is trained on
billions of words out of equivalent text in many different languages. For
example, it compares a German word like “Bible” versus a Russian word like
“Book”. Then it uses comparison results efficiently [18].
The translation of comprehension is still no perfect because the accuracy of
Google Translate varies with sentence, vocabulary complexity and by language.
2.3. The Role of Pedagogy in Learner Modeling
During the 1980's, computer scientists especially in AI continued to focus on the
problems of natural language, learner models, and deduction.
The pedagogical strategies are issued from psychological and didactic
research. It pertains with the underlying learning theories, i.e. behaviorism,
cognitivism and constructivism. These strategies are specified by a pedagogical
specialist – expert in education – and not a computer scientist [18].
The learner’s skill profile is recorded in a Learner Model. Jackson models
the learner’s behavior based on psychology of personality. The Expert modeler
monitors learner performance based on Jackson’s learning styles profiler.
In fact, pedagogy is a bridge in the gap between learning and modern
technologies. It caused ITS helps learner by “the best way”.
2.4. Jackson’s Learning Styles Profiler
Learning styles are various approaches or methods of the learning. An ITS is
based on accurate recognition of behaviors and individual characteristics. There
are some important factors in learning style such as aptitude, personality and
behavior [11].
It’s worth to be noted that the concept of “learning style” has emerged as a
focal point of much psychological research. In our proposed learner model, we
have adopted the Jackson’s learning styles profiler in order to model learning
styles of the learners. Jackson proposed five learning styles which are
*
http://translate.google.com/
7
summarized in Error! Reference source not found.4. Reader can find more
details in [2].
Table 4. Summarization of Learning Styles [8]
Learning Style Specifications
Sensation Seeking (SS) Believing that the experiences create learning.
Goal Oriented Achievers (GOA) Self-confident to achieve difficult and certain
target.
Emotionally Intelligent Achievers (EIA) Rational and goal-oriented.
Conscientious Achievers (CA) Responsible and insight creator.
Deep Learning Achievers (DLA) Interested in learning highly.
There are some approaches to model learner’s behavior. For example
Entistle approach [21] tries to connect the psychology concepts into the
effective variables on the learner’s view to ‘learning’. They are rarely applied in
adaptive education systems despite some benefits in education and psychology
of personality.
This section introduces Jackson model [22]. Jackson’s model is based on the
most recent researches in the psychology of personality. It also is an efficient
analyzer. So this paper just applies it among available models.
Jackson model has been proposed in Queensland University [22]. This
model has been developed by investigating personality, learning and evaluating
for fifteen years. The learning is based on new neurological psychology in this
model.
This model has some advantages. For example, it can export and report the
individual specifications of the learner. It is obtained on basis of the points
which the learner earns in the learning style (III-D).
The aim of this model is to recognize an ideal style based on pedagogy
principles which can model learner’s behavior and specifications in adaptive E-
Learning system. A key premise of Jackson’s model is that cognitive strategies
redirect Sensation Seeking to predict functional behaviors [8].
3. Proposed approach
ITSs build a model of the individual learner’s knowledge, difficulties and
misconceptions during interaction with the system. This learner model can be
compared with a model of the target domain. Suitable tutorial strategies can be
inferred by the system. They are appropriate for the learner according to the
contents of their learner model. In other words the educational interaction will be
matched with specific and unique needs of each learner [23].
8
A web-based ITSs is an E-Learning system based on web which can be used
ETEW. The first E-Learning system has been reported on 1995 [17, 24]. It is
web-based and intelligent. Learning of all courses is customized well at home via
web in this system. So learners can solve some examples and proper exercises
ETEW. Finally he can take a course exam virtually or physically.
A large number of ITSs have been conducted in a single language so far.
However, ITSs can support multiple languages with the integration of machine
translation such as Google Translate API among 64 languages.
3.1. Learning Environment
The learning environment is explained in this section. A learner can visits
website of intelligent virtual school after authentication. An intelligent Graphical
User Interface (GUI) is an interface between learners and intelligent tutor. This
section of system can affect learning efficiency. It should be user friendly as far
as possible.
An intelligent virtual class has some benefits such as graphical properties,
audio and video to make learning attractive. Moreover, some tools are available
to simplify learning process. Learners can communicate well with this inanimate
and non-physical system via these tools. Some facilities are Computer games,
frequently asked questions (FAQ), Video chat and email.
Various learners may be logged with different language according to
geographic region. Moreover all of E-learning systems can be customized for
different languages. For example, the same e-learning system can be run for
Iranian people with Persian contents.
This system has an Automatic Translator component which is able to
integrate language translatability with the system. Language modeling for each
learner is performed and store in the learner model. This component is
developed with a MultiTranslator such Google Translate API that provides free,
cheap and true translation in many languages at the same time.
Translation cost [25] and grammatical errors are some of the major
limitations in localizing e-learning system. They are reduced along with growing
of these systems.
This component helps learners to use the system without knowing any
foreign language. Some users can also improve their skills in second language.
3.2. Learning Process and Evaluation
This system uses a three layered structure to implement a concept: “Pre-test”,
“Learning concept” and “Post-test”.
9
The pre-test includes of some questions planned by an expert tutor in
advance. These questions can determine learner’s primary knowledge level. It
simulates learner model based on Jackson model [22]. The learning concept
depends on the learner level. So the best method is determined to train a learner.
A learning process starts up then. A post-test will evaluate the learner by some
questions finally.
Learning evaluation is the most important factor to determine learning
performance in E-Learning systems. It is performed with Pre-test and Post-test
layers of the proposed system. The evaluation has two levels, conceptual and
objective. The first one refers to the learner’s understanding out of the lesson
concept. The last one denotes to the learner’s understanding of the lesson topic.
The learner’s knowledge level is determined by concept level and objective level.
The tutor can extract proper questions from question-base collected by an
expert system, pre-test and post-test. He notes that a specific score is given to
each question.
Question selection should satisfy some rules:
1. None of them should be repetitious even if a learner would be trained a
specific concept several times.
2. The question must be planned for all sections of a concept entirely.
3. An expert tutor plans questions in all level.
The sequence, number and level of questions are determined according to learner
level and learning type intelligently. Sum of scores is calculated and then learner
level is determined after answering the questions. Error! Reference source not
found.5 presents five categories of learner’s knowledge level about a concept
[13, 14].
Table 5. Categories of Knowledge Level [8]
Knowledge Level Score
Excellent 86-100
Very good 71-85
Good 51-70
Average 31-50
The system modeler updates the learner’s model along with questions
answering. It can also save last academic status of learner and all his learning
records.
The proposed system integrates pedagogy and expert system with web-based
ITS structures. It develops a multilingual e-learning system by applying
automatic translator algorithms which is illustrated in Figure 2. So it can be used
10
by different nationalities. In addition of education application, it can be used as a
tool to learn a second language.
Figure 2. Block Diagram of Proposed System
4. Conclusion
This paper proposed a model for an adaptive, intelligent, multilingual and web-
based tutor. This model can models the learning styles of the learners using
Jackson’s learning styles profiler and expert system technology to enhance
learning performance. Previous E-Learning systems offer fixed languages and
multimedia web pages and facilities for user management and communication.
Based on the proposed model, E-Learning software is implemented on the
web which can identify learning styles, aptitude, characteristics and behaviors of
learner in order to provide appropriate individual learning content. On the other
hand, Jackson model can recognizes learner’s learning style. It leads to make
learner’s model closer. So the efficiency of adaptation process increases.
The proposed model acts as an intelligent tutor which can perform three
processes include of pre-test, learning concept and post-test based on learner’s
characteristic. This system uses expert simulator and its knowledge base as well.
Proposed model is also multilingual. So it helps learners to use the system
without knowing any foreign language. Some users can improve their second
language skills by this way.
11
A large number of ITS have been ever conducted in a single language. By
using the proposed model, ITSs can support multiple languages with the
integration of machine translation such as Google Translate API among 64
languages and 4032 language pairs.
The proposed system doesn’t have any drawback of previous system and
human expert tutor. It can improve the learning result significantly. In other
words, it helps learners to study in “the best way” using localized language.
References
1. K. Eustace, “Educational value of E-Learning in conventional and
complementary computing education,” in Proceedings of the 16th National
Advisory Committee on Computing Qualifications (NACCQ), Palmerston
North, New Zealand, 2003, pp. 53–62.
2. H. M. Ghadirli and M. Rastgarpour, “A Web-based Adaptive and Intelligent
Tutor by Expert Systems,” in The Second International Conference on
Advances in Computing and Information Technology (ACITY 2012),
Chennai, India, 2012.
3. P. Brusilovsky, “Intelligent tutoring systems for World-Wide Web,” in
Third International WWW Conference(Posters), Darmstadt, 1995, pp.42-45.
4. M. a. O. Specht, R., “ACE-Adaptive Courseware Environment,” The New
Review of Hypermedia and Multimedia, vol. 4, pp. 141-161, 1998.
5. P. Brusilovsky, “Methods and techniques of adaptive hypermedia,” User
Modeling and User-Adapted Interaction, vol. 6, pp. 87-129, 1996.
6. S. Ohlsson, “Some Principles of Intelligent Tutoring,” In Lawler & Yazdani
(Eds.), Artificial Intelligence and Education, vol. 1, Ablex: Norwood, NJ,
pp. 203-238, 1987.
7. J. Yang and E. Lange, “SYSTRAN on AltaVista: A user study on real-time
machine translation on the Internet,” Proceedings of the 3rd Conference of
the Association for Machine Translation in the Americas, Langhorne, pp.
275-285, 1998.
8. H. M. Ghadirli and M. Rastgarpour, “A Model for an Intelligent and
Adaptive Tutor based on Web by Jackson’s Learning Styles Profiler and
Expert Systems,” Lecture Notes in Engineering and Computer Science:
Proceedings of The International MultiConference of Engineers and
Computer Scientists 2012, IMECS 2012, 14-16 March, 2012, Hong Kong,
pp 63-67.
9. E. Shaw, et al., “Pedagogical agents on the web,” in Proceedings Of Third
International Conference on Autonomous Agents, 1999, pp. 283-290.
10. Shute, V.J. and Regian, J.W. (1990). “Rose Garden Promises of Intelligent
Tutoring Systems: Blossom or Thorn?,” Presented at Space Operations,
Automation and Robotics Conference, June 1990, Albuquerque, NM.
12
11. Koedinger, K., & Anderson, J. (1995). “Intelligent tutoring goes to the big
city,” Proceedings of the International Conference on Artificial Intelligence
in Education, Jim Greer (Ed). AACE: Charlottesville, VA pp. 421-428.
12. C. Jackson, Learning Styles and its measurement: An applied
neuropsychological model of learning for business and education: UK: PSi-
Press, 2002.
13. Weber, G., & Brusilovsky, P. (2001). “ELM-ART an adaptive versatile
system for web-based instruction,” International Journal of Artificial
Intelligence and Education, this volume.
14. L. Kettel, J. Thomson, and J. Greer, “Generating individualized hypermedia
applications,” In Proceedings of ITS-2000 workshop on adaptive and
intelligent webbased education systems, 2000.
15. N. Entwistle, Styles of learning and teaching: an integrated outline of
educational psychology for students, teachers, and lecturers: David Fulton
Publish, 1989.
16. P. Brusilovsky, “Adaptive and intelligent technologies for web-based
eduction,” KI, vol. 13, pp. 19-25, 1999.
17. Bra PD, Santic T, Brusilovsky P. AHA! meets Interbook, and more.
Proceedings of the World Conference on E-Learning, Phoenix, AZ,
November 2003, Rossett A (ed.). AACE: Norfolk, VA, 2003; 57–64.
18. M. Aiken, et al., “Automatic translation in multilingual electronic meetings,
” Translation Journal, vol. 13, 2009.
19. M. Siadaty and F. Taghiyareh, “PALS2: Pedagogically adaptive learning
system based on learning styles, ” 2007, pp. 616-618.
20. C. Yang, “An Expert System for Collision Avoidance and Its Application,”
PH.D. Thesis, 1995.
21. J. M. Ragusa, “The Synergistic Integration of Expert Systems and
Multimedia within an Intelligent Computer Aided environmental tutoring
system,” in Proceedings of the 3rd world congress on expert systems, 1960.
22. A. S. Gertner, VanLehn, K., “ANDES: A Coached Problem-Solving
Environment for Physics. In Intelligent Tutoring Systems,” in Fifth
International Conference, ITS New York, 2000, pp. 133–142.
23. Y. Cui and S. Bull, “Context and learner modelling for the mobile foreign
language learner,” System, vol. 33, pp. 353-367, 2005.
24. K. F. VanLehn, R.; Jordan, P.; Murray, C.; Osan, R.Ringenberg, M.; Rosé,
C. P.; Schulze, K.; Shelby, R.; Treacy, D.; Weinstein, A.; and Wintersgill,,
“Fading and Deepening: The Next Steps for ANDES and Other Model-
Tracing Tutors,” Intelligent Tutoring Systems, 2000.
25. B. L. Hafsa and E. M. des Ingénieurs, “E-learning globalization in
multilingual and multicultural environment,” WSEAS Proceedings NNA-
FSFS-EC, pp. 29-31, 2003.
