An adaptive and cooperative telelearning system: SMART-learning.
In this article, we will discuss the adaptative and cooperative aspects that are the principal characteristics of the telelearning system, SMART-Learning. First, we present the telelearning problems to which our system tries to answer. Then, we present the conceptual approach that allows us to define the pedagogical needs, making it possible for learners to follow a course that fits their aptitudes and objectives. Following this, we introduce the various concepts and components of the SMART-Learning architecture that allows set-up to adaptative telelearning. Next, we show the main stages of the learning process, giving access to a course with several profiles. Finally, we briefly present the production process of courses in conformity with the concepts previously defined.
First, we will point out the main motivations that guide the current evolution of telelearning. We consider that these motivations are fundamental, and it is necessary to keep them in mind to answer to them throughout our system design:
1. The demand for training is increasing. In particular, the greatest needs lie in continuing education field and in those fields that evolve quickly, such as new information technologies. Pedagogical resources are usually insufficient or under-exploited -- for example, an expert teacher for small classes.
2. Learners have space and time constraints, especially in continuing education.
3. Learners have different learning rates. The courses currently given are not adapted to this situation. Some learners can be bored in a course because they understand it more quickly than their classmates. Others will be quickly overwhelmed because they didn't have enough explanations.
4. The courses currently given do not sufficiently take into account technological progress. These courses integrate few multimedia supports that improve the appearance of the courses and make them more attractive to the learner.
In this article, we develop the aspects of our system primarily related to the first three points.
Our telelearning system is based on an asynchronous mode without exclusion of synchronous sessions. The asynchronous mode makes it possible to provide courses without time or space constraints, while the synchronous mode allows an open debate in real time.
Even synchronous telelearning allows the chance to partly answer the strong demand for training and avoid moving. It forces learners to attend the course at the same moment and for the same duration. In SMART-Learning, telelearning constitutes a complementary tool to the asynchronous mode, allowing direct communication between teachers and learners.
Several asynchronous telelearning applications were developed to produce and provide courses on a network environment, in particular, the Internet. The most well-known applications are TopClass and WebCT (2002) for commercial applications, and Virtual-U (Harasim, 1999) and MANIC (Padhye & Kurose, 1999) for research applications.
In these various systems, learners access the system without time constraint at the moment that is best appropriate for them. However, the pedagogical contents presented to them are the same for all other learners. They are then obliged to swallow information that can seem obvious or incomprehensible.
Moreover, courses that are generally provided on the Internet are simple hypertext documents. The permissive structure of hypertext and the reading methods to which this structure leads are contrary to some pedagogical principles. Learners can access various parts of a course without the course taking their aptitudes into account, nor the order that is necessary to follow between the parts. Consequently, the pedagogical objectives -- specialty, training type -- pursued by the access to this course will not be achieved.
Generally, in existing systems, learners are left alone in front of their individual courses. Cooperation is necessary to provide a pedagogical support for them. This cooperation -- with the teacher or other learners, if it exists in some systems -- is done in a traditional way. It does not lead to a pedagogical control on the progression of the learner in the course.
Generally, the ITS (Intelligent Tutorial Systems) that are interested in the adaptability of the learning material are based on static techniques of generation (Brusilovsky, 2000; La Teja, Longpre, & Paquette, 2000) suggested by the author and chosen by the learner, one for all without taking into account the feedback of the learner with the system during the process if training. Thus, it would be desirable to develop an open or generic system that allows a dynamic adaptability. We make an effort in SMARTLearning to provide solutions for the described problems.
To benefit more from the advantages of asynchronous telelearning, we believe that a personalized course must be provided to the learners. A personalized course must contain pedagogical material adapted to the learners relative to individual abilities, as well as the objectives learners wish to reach by this training. They will then spend more time in front of the learning station, accessing the course independent of the time, date or duration. Time usage will then be optimized. This characteristic is even more appreciable in continuing education, where the time and space constraints are stronger due to work and employer commitments (Martin, 1997).
Moreover, it is not acceptable to leave the learners' access to training contents in an anarchistic manner without being concerned with their abilities and interests to access to a given part of the course. This observation is important, as the objective announced by some pedagogical specialists is to also provide training leading to telelearning degrees to answer to the current strong demand for training in some fields.
To resolve these problems while respecting the fundamental telelearning pedagogical principles (Martin, 1997; Bloom, Krathwohl, & Masia, 1970), and particularly, to provide an active learning environment (Kannat, 1997), all learners must have their own course, a specific course that is adapted to their aptitudes and objectives.
The most used method to meet this need is a course under several versions. Each version corresponds to a specific learner profile taking into account individual abilities and training objectives. For this rigid solution, adding a new profile or modifying a parameter of the profile requires the addition of another version.
To avoid the multiplicity of the same course, SMART-Learning proposes an approach based on the production of a generic course. This approach is an intermediate solution between the directive teaching of a rigid version and the hypertexts that do not present any control on the course learning process. This approach makes it possible to generate an individual course for all learners based on their profiles by using the generation system (Figure 1). If the profile is further modified to an interaction with the system -- an evaluation -- for example, the course will be automatically regenerated to take account of the new knowledge obtained by the learner.
Before beginning a course, the system needs to know the learner to build the profile. The profile is then used to present the organization and the content of the course to the learner. However, this means that the author of the course must parameterize the various parts of the course according to the objectives pursued, as well as the prerequisites necessary for reaching each specific part.
Two key elements must be taken into account: the structure of the generic course according to the educational objectives, and the learner profile that will help to validate the choice of a course or parts of a course to be presented to the learner.
The learner profile
The learner's profile is a key element in the SMART-Learning education process, as it intervenes on all the learning process levels. It is represented by some information that characterizes one learner in this process (PAPI, 2000). The profile consists of two types of information:
1. General information relative to the learner: This information, in general, does not change a lot during the learning process. It is made up of the identity of the learners, the learning language[s], the objectives to be reached, (diploma, degree of knowledge), the training level to be reached (engineer, technician, manager) and the field of knowledge wished to be reached in the field (computer, networks, telecom, management); and
2. Contextual information: This information is dynamic. It changes with the evolution of the learner in the process of learning, as well as with a better knowledge of the learner by the system. It corresponds, for example, to the learner's training level so as to control progression. The results of examinations and other tests feed the profile continuously. They allow, with the reaction speed of the learner, to better know the learners' psychological level and factors taking into account the teaching contents to be preserved and in the duration with that it is adequate to present it.
For the same training objectives, the learners will have their own course corresponding to their language, learning speed or capacity. This method makes it possible to obtain a targeted teaching and better adapted. It will then be more interesting than a traditional course and more pedagogical than a gross document hypertext.
Further, we can say, that at every moment, the individual profiles reflect a reliable and complete intellectual image of the learner and keeps tracks of progress. This way, the next access will take account of the real profile and provide adequate knowledge of the learning.
The generic course
The generic course is intended for all the profiles. It is a document that contains, in addition to the learning material intended for the learners, the access conditions to each part of the course so that the system can generate a personalized course for each learner. A generic course must take into account the various objectives as well as the various profiles of the learners who will have access to this course. It is necessary to account for all of the details, even the optional parts of the course. In a traditional setting, the teacher is present during the training session and can add missing information or make a reminder not initially considered. In the case of telelearning, if the contents of the reminder are not initially considered, teaching will be incomplete.
On one hand, this method may seem complex and even useless when a course is intended for only one profile, though it is extremely rare to find a class, even the most homogeneous, with learners having exactly the same levels.
On the other hand, even if the course preparation is complex, it will allow to minimize the efforts and save time when the course is intended for learners with several profiles compared with the preparation of several versions of the same course.
The idea, then, is to keep the same structure for the course and to change for each profile only the part of the contents that relates to it.
For example, for a course intended for both engineers and technicians, the shared contents of the course remain unchanged, but to change specific parts of the course would be sufficient.
This method also makes it possible to avoid errors due to course updates on different course versions intended for different profiles. For example, an update can be carried out on the version intended for the engineers and not for technicians.
In the same way, for a course intended for a public of various languages -- currently a top need -- there can be differences of structure and contents in the versions corresponding to each language.
In SMART-Learning, the generic course is a single course. It will keep the same exact structure that will also help carry out methodical and systematic updates.
GENERIC COURSE MODEL
It is important to have a basic model for all the courses of SMART-Learning to:
1. take account of all the parameters of the learners profiles;
2. take account of the various pedagogical constraints that are defined in the model; and
3. help the author to conceive a course based on a model in conformity with a preset teaching method.
During the course production phase, the author must conform to the course model that will represent a course mould that guides the author in the realization of the course.
The course model can also relate a teaching methodology. The model will guide the author to better respect the methodology while keeping a personalized aspect for the course.
Consequently, the modeling of a course is carried out according to two following aspects.
General structure of the course
The general course structure describes aspects related to data organization in a course. The course, thus built, must be sufficiently flexible to be appropriate for a broad variety of learners to meet the training needs and the pedagogical objectives. Each course production is based on a pedagogical diagram that is used to structure the course contents to facilitate learning.
The generic course is a set of pedagogical sequences structured as a tree with a single root. Each sequence constitutes a distinct whole. It can be a course, a chapter, an evaluation and so forth. A pedagogical sequence is broken up into other pedagogical sequences. The sequences of the lowest granularity are called elementary pedagogical sequences (Ajhoun & Benkiran, 2000b). This arborescent structuring makes it possible to be able to adapt a course to several pedagogical methods and the structuring of courses that they induce (Figure 2).
The elementary pedagogical sequence is then subdivided into media elements -- text, sound, images and so forth. The access to a pedagogical sequence, or even to a media element, depends on the profile parameters of the learner for whom the course is intended. The granularity of the access selection is left to the authors according to their course contents.
For example, for the same course objectives -- degree, specialty, field and so forth -- several levels of learner ability can exist. This makes it possible to provide a course that is adaptable to individual abilities. The teaching language can convey the course. In our system, learners can choose the language that is most suitable for a better assimilation of the course. Then, the learners access the adequate pedagogical sequences and media elements.
The pedagogical progression graph
Once the course is generated, it begins with the first pedagogical sequence of the course. At the end of a sequence, an interaction by the learner, an evaluation for example, can happen to test acquired knowledge. According to the results of the test, the learner will be authorized to pass to the following sequence or remake the same sequence, if necessary, with a modified profile taking account of re-examined abilities. Passing from one sequence to another can also be conditioned with an event coming from the teacher. These interactions and events or conditions correspond to a direct or indirect cooperation through the course between the teacher and the learners.
Thus, the passage from one pedagogical sequence to another is not carried out in a random or automatic way, but performed according to precise pedagogical rules. These rules are translated by relations between the pedagogical sequences, and it is the course author's task to clearly define these transitions in the form of a pedagogical progression graph or network (Figure 3). In fact, the progression in the course is based on a set of transitions between the pedagogical sequences of the course until the last sequence of the course is reached.
The pedagogical progression graph represents all of the pedagogical rules (Ajhoun & Benkiran, 2000c; Ajhoun, 2001). It controls the succession of various sequences of the course to achieve a pedagogical goal (succession, choice).
In Figure 3, at the end of the pedagogical sequence PS1, the learner must meet a condition -- pass a test, for example. In this case, the learner can pass to the next pedagogical sequence, either PS2 or PS4, these sequences being independent of each other. The learner cannot, however, reach PS5 until the completion of sequences PS2 and PS3 in that order, or the completion of PS4.
If condition PS1 is not met, the learner must repeat PS1 with a consequently modified profile leading, for example, to differentiate the visit duration of PS1.
PRODUCTION AND LEARNING PROCESS
The courses are stored in the form of hypermedia documents in servers accessible by the learners through the network. The learners can access courses relative to their profile. During the learning process (Figure 4), the system generates an individual and personalized course, specific from the generic course and the learner profile (Benkiran, Ajhoun, & Belqasmi, 2000).
The course must then be produced to be used within the learning process defined above. The author has to respect the model defined in the preceding section. Also, he must produce a well-structured course based on the model and the pedagogical material to be presented to the learner according to the learner profile (Figure 5).
During this phase, the conformity of the course compared to the model will be checked, in such a way that the author will be sure that the course contains the main elements and that adaptability was suitably taken into account.
Once the specific course is generated, the learner will access the course pedagogical sequences in an interactive way. Any significant interaction by the learner that modifies the profile can require either a regeneration or modification of the course or a specific pedagogical sequence to take account of the progress in the course. The modification will be carried out in accordance with the pedagogical graph, to take account of the relations between the different pedagogical sequences of the course.
A significant interaction can correspond, for example, to the result of an evaluation, with other parameters such as the answer speed, authorizing the passage to another pedagogical sequence of the course in case of the parameters are satisfied.
One can think that the regeneration of the course is an expensive task in processing and time. To simplify this task, we chose a simpler and more powerful technique. At SMART-Learning level, the problem is solved in a judicious way by separating (Figure 4):
1. the content part of the course (structure and media elements); and,
2. the selection conditions of the course elements based on the profile.
Thus, only the selection conditions will be modified according to the profile changes, and not the generic course. Then, course elements corresponding to the modified selection conditions will be presented to the user.
It is now a question of finding a means to describe a generic course with its content part and its presentation conditions part.
Moreover, making our system accessible to a large number of users requires the choice of means and tools that are available and known by a great number of users. In our view, to provide courses to learners anywhere in the world, the use of the Internet is impossible to circumvent, as the number of users of Internet increases continuously.
XML, eXtensible Markup Language (Rusty, 2000; Hunter 2000), famous for its ease of exchange of complex documents on the Internet, was selected to structure the courses. We can use XML in a simple way to solve the adaptability problem in its realization phase. We thus defined an approach based on tools of the XML family (DTD, XSL, DOM).
As the course is produced once and the system generates several versions, the functional model consists of two main components:
1. one component for the course production -- generic course and pedagogical progression graph; and
2. the other component for the learning process -- specific course with profile modification (Figure 4).
The rest of the course's structure is easily carried out with a DTD document (Document Type Definition). To structure the course, the DTD imposes different construction elements as well as the links between these elements. The use of the DTD ensures coherence of the course structure.
XML documents allow the description of the course contents without worrying about presentation problems. The style sheets describe the presentation form independently. We, thus, separate the pedagogical material and the form in two distinct documents. This makes it possible to easily modify the presentation form of a course. XSL (eXtensible Stylesheet Language) is the description language that deals with adaptability problems and allows the expression of the pedagogical selection conditions of the course sequences thanks to its abilities of transformation and conditional presentation (Rusty, 2000).
Thus, any profile modification of a learner will lead to a modification of the corresponding XSL document and the presentation conditions to present the course contents to the learner that correspond to the new profile.
The course adaptability is, thus, solved in a simple way while respecting the pedagogical constraints and remaining in conformity with the presentation rules on the Web.
The realization of a prototype enabled us to confirm the relevance of our conceptual choices, the simplicity and the performances of the selected method to implement multimedia adaptative courses for various learner profiles by using generic courses based on hypertext documents. The use of XML/XSL to provide adaptative courses on the Internet has several advantages. We emphasize its simplicity, the low data transfer rate between the learner and the course server and the adaptability of the course to each learner.
The authors would like to thank Mr. Mounir Mokhtari for reading early drafts and providing suggestions about English style.
Ajhoun, R. (2001). Adaptabilite et cooperation des cours dans le systeme de tele-enseignement SMART-Learning. These d'etat (publication of doctoral theses), pp. 105-146. Rabat, Morocco: Ecole Mohammadia d'lngenieurs.
Ajhoun, R., & Benkiran, A. (2000). What tools for adaptative multimedia telelearning. Proceedings of ED-MEDIA 2000, World Conference on Educational Multimedia, Hypermedia and Telecommunications (pp. 1202-1203), Montreal, Canada. Charlottesville, VA: Association for the Advancement of Computing in Education.
Ajhoun, R., & Benkiran, A. (2000b). A SMART-Learning: Adaptative telelearning system on the Web. Proceedings of TICE 2000, Colloque international sur les Technologies de l'information et de la communication dans les Enseignements d'lngenieurs et dans l'Industrie (pp. 23-29), Troyes, France.
Ajhoun, R., & Benkiran, A. (2000c). Modeling adaptative courses for telelearning. Proceedings of ITHET 2000, the lnternational Conference on Information Technology Based Higher Education and Training (pp. 271-275), Istanbul, Turkey.
Benkiran, A., & Ajhoun, R. (2000). SMART Learning: Systeme pour un tele-enseignement adaptatif et cooperatif. Vlliemes Journees Internationales de Technologie, 2(5), 91-100.
Benkiran, A., Ajhoun, R., & Belqasmi, Y. (2000). Adaptability of multimedia courses using intelligent hypertexts. Proceedings of TechEd2000: Technology in Education International Conference (pp. 7-13), Palm Springs, CA.
Bloom, B.S, Krathwohl, DR., & Masia, B.B. (1970). Taxonomie des objectifs pedagogiques. Domaine affectif Education nouvelle (Tome 2).
Brusilovsky, P. (2000). Adaptive educational systems on the World Wide Web: A review of available technologies. Retrieved from the World Wide Web on University of Massachusetts web-site on April 12, 2002 from http://www-aml.cs.umass.edu/~stern/webits/itsworkshop/brusilovsky.htm l
Harasim, L. (1999). A framework for online learning: The Virtual-U. IEEE Computer, 32(9), 44-49.
Hunter, D. (2000). Initiation a XML avec trois eutdes de cas detaillees. Paris: Edition Eyrolles.
Kannat, S.E. (1997). Etude de faisabilite au niveau reseau de l'application de tele-enseignement: Specification, elaboration et mise en ceuvre. Internal Report, Institut de'Informatique et Methemetiques Appliquees de Grenoble. Retrieved from the World Wide Web on April 22, 2002 from http://www-lsr.imag.fr.
La Teja, I., Longpre, A., & Paquette, G. (2000). Designing adaptable environments for the Web: A case study. In Proceedings of ED-MEDIA 2000, World Conference on Educational Multimedia, Hypermedia and Telecommunications (pp. 243-248), Montreal, Canada. Charlottesville, VA: Association for the Advancement of Computing in Education.
Martin, P. (1997). La conception pedagogique de systemes d'apprentissage multimedia interactif: Fondements, methodologie et problematique, 1(3). Directeur du groupe de recherche GRAIM, Faculte des sciences de I'education, Universite Laval.
Padhye, J., & Kurose, J. (1999). Continuous-media courseware server: A study of client interactions. IEEE Internet Computing, 3(2), 65-73.
PAPI (2000). Specification learning technology: Public and private information, 5.0. Available: http://ltsc.ieee.org/
Rusty, H. (2000). XML le guide de l'utilisateur, Paris: Osman Eyrolles Multimedia.
WebCT (2002). [Website] Retrieved from the World Wide Web on April 22, 2002 from http://www.webct.com.
M. AMINE BENKIRAN, UFR RESEAUX INFORMATIQUE ET MULTIMEDIA, ECOLE MOHAMMADIA D'INGENIEURS (EMI), RABAT, MOROCCO
RACHIDA AJHOUN, ECOLE NATIONALE SUPERIEURE D', INFORMATIQUE ET D', ANALYSES DES SYSTEMES (ENSIAS) AND UFR RESEAUX INFORMATIQUE ET MULTIMEDIA, RABAT, MOROCCO
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||System for Multimedia Adaptative and coopeRative Telelearning|
|Publication:||International Journal on E-Learning|
|Date:||Apr 1, 2002|
|Previous Article:||A constructivist approach to web course design: a review of the literature.|
|Next Article:||WebNet Journal & International Journal on Educational Telecommunications evolve into the International Journal on E-Learning. (Editor's Note).|