The goal of SMART-Learning is to develop a telelearning system where learners are able to follow courses at their own pace and access only those elements of the course that relate to them. Production is based on intelligent hypertexts (e.g., XML). These hypertexts, which allow diffusion of the courses on the Internet, introduce simplicity for adapting the course to the learner. The power of our system is the use of widely available tools. This availability will guarantee a large opening and a larger audience.
In this proposal, we will highlight the principal needs for carrying out an adaptative training program as well as adequate tools for development.
Over the last several years, the field of telecommunication and multimedia data processing have seen visible progress. Because researchers in the field of telelearning have exploited these new technologies, it is possible to improve teaching methods in order to satisfy the needs of users of this kind of application.
SMART-Learning, or System for Multimedia Adaptative and coopeRative Telelearning, is a project at the Réseaux Informatiques et Multimédia (RIM) research unit at EMI. The fundamental goal of this project is to develop a virtual university based on the asynchronous mode without excluding synchronous sessions. Indeed, the asynchronous mode is famous for its great ability to provide courses without temporal constraints, whereas the synchronous mode allows a better open debate in real time. In addition, the fundamental goal is adapting courses to the learner according to teaching requirements and learner capacities.
In the SMART-Learning system, the learners are at the center of the design process. The main pedagogical characteristics of our approach are summarized by the following points :
This paper is organized in two sections. The first section presents the design and the different components of our architecture. The second section is reserved for the techniques and methods for the adaptative aspect of the courses.
Much educational research deals with styles of learning and how technology can be used to support effective learning. The challenge is to (1) choose the best theory applicable to telelearning and (2) use the most appropriate technology.
Our approach makes a significant advance when compared with the typical approach, which provides courses as hypertext documents . In the typical approach, a learner can access any sequence of a course without order constraints on these sequences and without having the necessary knowledge. Consequently, the teaching objectives (e.g., specialty, type of training) aimed by the access to the course will not be reached. In our view and to ensure a pedagogical learning, the course must take account of the learner profile (e.g., capacities of training, previous knowledge, language) and the learner evolution (e.g., prerequisite, speed).
For this reason, we fixed as a strategic objective the development of an adaptative multimedia telelearning application. To achieve this goal, several aspects were taken into account during the steps in our application design.
To fulfill our strategic objectives, the main components of our application are the same as those of a traditional university. Therefore, the basic model of our architecture is constituted from three main components: administration, production, and learning (figure 1).
To make the courses evolutionary, the author uses the interactions of learners with the system (e.g., questions, remarks, badly understood sequences) to improve the course. Using this feedback, the author can modify the contents and improve the teaching techniques if necessary. This significant aspect is translated into the architecture of our application by a close dependence between the production and training phases.
Figure 1: A generic architecture for SMART-Learning
Administration is an important component because at this stage information related to the participants is inputted. The following points summarize the main tasks of this component :
The author is responsible for building a course based on an educational model that respects the teaching objectives. The model defines the hierarchy of the course and identifies the presentation structure by indicating the relationship between the various elements of the course in a pedagogical manner. Techniques for improving the contents of the course or the pedagogical method used are taken into account in this component. The course's structure as well as its contained elements are stored in a database.
This component provides the individual course to the learner . The learning process comprises two main modules: a teacher module and a learner module. The teacher module constitutes the engine of the process of learning; it acts as an assistant (controller) for the learner. Its responsibilities include motivating and encouraging learners remotely, maintaining discipline, helping learners prepare for a course (e.g., showing learners how to print a diagram that will help them understand the course), answering questions, evaluating learners, and programming synchronous meetings. Also in our architecture, we will integrate synchronous and asynchronous tools of communication that are available, like electronic mail (e-mail), visioconferences, and chats among the various participants (e.g., learners, teachers, author, administrator) to meet the traditional needs in a university. The learner module is responsible for presenting adequate sequences to learners according to their skills and objectives.
For teaching purposes, cooperative learning is more interesting than individual training. This is why in our design we propose the concept of the virtual class to which are registered the common points (e.g., the same follow-up of course period, the same specialty, and so on). Taking advantage of the concept of the classroom, the course itself will be enriched by learners' remarks and questions. For the same reason, the synchronous meetings organized by the teacher will be for learners from the same class.
Making our system accessible to a large number of users requires choosing the means and tools that are available and familiar to the greatest number of users. In our view, to provide courses to learners anywhere in the world, it is impossible to circumvent the use of the Internet. Indeed, the number of Internet users increases continuously. We notice that choosing the Internet for our application does not exclude the use of other more powerful networks - for example, the Integrated Services Digital Network (ISDN) -- or of less powerful networks (e.g., telephone networks).
Designing software for the World Wide Web poses particular challenges to the design of user interfaces, in part because of (1) the limitation of bandwidth and (2) the very nature of hypertext transfer protocol (HTTP) browsers as document retrieval systems  .
The Web is the most widely used technology, and the associated tools are recognized to be powerful and widely available. This naturally explains why we have chosen to use the Web. We are aware that our choice of the Internet, and the Web, has many drawbacks, but we think that the advantages prevail over the disadvantages.
The Web was conceived so that users could navigate and seek information on the Internet; the Web was not designed for educational uses. Its main disadvantages are the permissive structure of the HTML (which is opposed to some educational principles), the bad interactivity, the nonguaranteed quality of service, and the difficult synchronization between media. These difficulties led us to find solutions at the design level in order to avoid these problems.
The flow rate and response time of access to the Internet network depend mainly on the time and the place of access. Transferring the elements of a course, which are multimedia documents, can result in bad performance, in particular in the southern countries. To circumvent this disadvantage, we propose several access modes. One is the off-line mode, which exists in fact to store all the media elements of the course coded (for security reasons) locally. All the process of training, like educational rules, is done remotely. To show a media element of the course to the learner, access is done locally (e.g., with a CD-ROM or disk) instead of transferring the information from the server. This makes it possible to improve the speed of the presentation and the interactivity with the system.
In this section, we are mainly interested in presenting a system based on powerful techniques that allow courses to be reliably adapted to learners, using tools that are widely available and easy to use. Our approach is based on producing a generic course  , which will be used to generate, at learning process, several specific courses according to learner profiles.
This approach is an intermediate step between the directing whole (equivalent to the method of several versions, each version for one profile) and the hypertext, which does not present any control over learners' progress in the course.
This approach makes it possible to generate an individual course for each learner based on his or her profile by using a system of generation (figure 2). If the profile of a learner is modified as a result of an evaluation, for example, this course will be automatically regenerated in order to take into account the new knowledge acquired by the learner.
Figure 2: Functional diagram of the learning process
The author of the generic course must parameterize the various sequences of the course according to the objectives pursued and the prerequisites necessary for reaching a specific educational sequence   .
Two key elements must be taken into account: (1) the modeling of the generic courses according to the educational objectives and (2) the learner profile, which will help to validate the choice of a course or sequence of courses for the learner.
The learner profile is the key element in the SMART-Learning education process. It is the information set that characterizes a learner and is useful for the learning process. The main information includes capacities, learning objectives, psychological factors, and others  .
Learners with the same capacities will have a course tailored to his or her psychological factors and learning objectives. This makes it possible to have targeted and better adapted learning, which is more interesting than traditional teaching and more educational than rough hypertext.
Further, we can say that at every moment, the profile reflects a reliable and complete intellectual image of the learner. The profile keeps track of the learner's progress in the course so that subsequent accesses will take into account the updated profile, providing adequate knowledge of the learner.
A generic course is a document that contains (1) the learning material to be given to learners and (2) the access conditions to each sequence of the course, according to the course's various objectives and the types of learner profiles that will have access to it.
We believe it is necessary to model the generic courses because of two reasons:
Our modeling is done in two phases: the general structure and the educational graph.
In this phase we will present some aspects related to the organization of the material that allows for the construction of the courses. The generic course must be sufficiently flexible to fit a large variety of learners in order to adequately meet training needs according to educational objectives. The production of any course is based on an educational design that is used to structure the contents in order to facilitate learning  .
The course is a set of educational sequences; each educational sequence makes a distinct whole. Access to a sequence depends on the objective (e.g., specialty, type of training). For the same objective, we can have several types of learner capacities. This makes it possible to provide a course to a learner according to his or her capacity. Also, the language the in which the course is taught affects understanding. In our model, the learner can choose his or her preferred language in order to better understand the course.
We now wish to generate an individual course for a given learner. Once the course is generated, the learner starts the first part. At the end of this part, the system must take into account the learner interactions during this educational sequence (e.g., evaluation). The student will then have access to a specific educational sequence of the course or will redo the same educational sequence, if necessary, with a modified profile (figure 3). The transition from one educational sequence to another is therefore not made in a random or automatic manner, but according to an educational approach. This approach is based on the relationships between the sequences. It is the author of the course who clearly defines these relationships in the form of an educational graph. Finally, the course consists of a multitude of transitions between the various sequences in a positive and progressive way.
Figure 3: A simple graph of an Educational Network
The remainder of this article presents the methods that we conceived in order to produce a course that respects the model. We also discuss the techniques used to make learners follow courses on the Internet according to their individual profiles.
The author of a course must obviously respect the model defined in the previous section. He or she will be able to produce a well-structured course based on this model and the learning material to be given to learners according to their profiles.
The (Language) language , reputed by its facility to exchange complex documents on the Internet, have been selected to be able to structure the course. To solve the problem of adaptability in its practical phase, we have defined an approach based on some XML-specific tools.
Because the course is produced only once and the system generates specific versions, our functional model is constituted by two components -- one for the production of the courses and the other for the learning process (figure 4).
The course modeling is easily made by the Document Type Definition (DTD)document. To structure the course, the DTD imposes the various constituting elements and the hierarchy between them. The use of the DTD ensures the coherence of the course's structure.
An XML document makes it possible to describe the contents of the course without being concerned with its appearance. It is the style sheet that changes to describe the way in which the document XML is displayed. This permits us to easily modify the way in which a course is displayed and to add simply information. The style sheet language we choose is eXtensible Stylesheet Language (XSL) due to its strong capacities compared with other languages, such as CCS .
In order to use all the facilities provided by XML or the tools that are based on XML, and with the aim of optimizing the processing to provide learners with specific courses tailored to their profiles, we distinguished four different documents that the author of the course produces (figure 5):
Separating the course into several documents makes it possible to improve the extensibility of the course and to facilitate improvement of the sequences. Furthermore, generic XML documents and generic XSL documents can be useful for the production of several courses, by modifying the sequences.
After the various documents of the course are produced, they are stored in an external peripheral that learners can access through the Network according to their profiles. At the learning process (figure 4), our system generates an individual XSL document for the learner using the generic XSL document and the learner profile. The combination of the generic XML document and individual XSL document will, of course, constitute the individual course.
At the end of each educational sequence, and if necessary, learners make a self-evaluation test to control the knowledge obtained during this part. According to the result and other parameters (e.g., speed) and based on the educational graph, the system chooses the next educational sequence to present to the learner. The modification of the profile, following an evaluation, automatically regenerates a new individual XSL document.
Figure 4: Functional Model
To test our design and evaluate our model, we produced a prototype. The general architecture is described in figure 5. In addition to the component of users administration, we carried out the techniques of adaptability that we had tested on a demo course. All the communications between learners and the server is based on HTTP protocol. On the other hand, our implementation is carried out in the Java environment for the high level of portability it provides and the set of tools available (e.g., security manager, network package) .
To access a course, the learner sends a request that contains his or her personal information and curriculum vitae. This registration will be valid if the learner has all the knowledge necessary to start this course. The system generates a learner profile based on this information.
Figure 5: Prototype Architecture
The course document is constituted with an XML and XSL document. The XML document is generated from a prototype DTD document, which we had realized, and the XML entities are composed from text, images, and audio files.
The system comprises three main components:
The prototype of our system was carried out and tested at our research unit with great success and satisfaction. Furthermore, we underline the use of XML/XSL, which has several advantages, to provide adaptable courses on the Internet. We mention its simplicity, its ability to transfer the structure of the individual course from the course server to the learners, the way it allows the uploading of only those elements the learner needs, and how one can anticipate the transfer of the course elements that will be used later. Another advantage lies in the fact that our application requires only one light processing to adapt the course to one learner.
 Mahamane Ato A. ss/dir A. Benkiran Mémoire de DESA (Diplôme des Etudes Supérieurs Approfondies) - Système de Télé-enseignement asynchrone: conception et réalisation de cours adaptatifs, Ecole Mohammadia d'Ingénieurs, Octobre 1999.
 Christian Pantel: A framework for comparative We-based learning environments, October 1997, Simon Fraser University, thesis.
 Padhye J., Kurose J. Continuous-Media Courseware Server: A Study of Client Interactions IEEE Internet Computing, Vol. 3, No. 2, March/April 1999.
 Benkiran A.,Ajhoun R., Belqasmi Y. SMART-Learning: Système pour un télé-enseignement adaptatif et coopératif VIIiemes journées internationales de technologie, Beyrouth, Liban, Mai 2000.
 Ajhoun R., Benkiran A., Belqasmi Y. A Methodology for Adaptative Courses on the Web, proceedings of DEC'2000 the 7th Distance Education Conference, Austin, Texas, January 2000.
 Benkiran A., Ajhoun R., Belqasmi Y. Adaptability of Multimédia Courses using Intelligent Hypertexts, proceedings of TechEd2000 - Technology in Education International Conference, Palm Springs, California, March 2000.
 Khalifa M. Effects of Hypertext on Knowledge Construction Department of Information Systems. City University of Hong Kong, IEEE 1998.
 Abunawass A., Neumann K., Maskarinec M. Heuristics for the design of an effective distance learning curriculum proceeding of the IASTED International Conference, Mexico,1998.
 Jonas A. Montilvac C. Designing Web based study guides for distance education courses proceeding of the IASTED International Conference, Mexico,1998.
 Bloom B.S, Krathwohl D.R,.Masia B.B. Taxonomie des objectifs pédagogiques. Tome 2. domaine affectif Education nouvelle, Inc. 1970.
 Fleury M. Implications de certains principes de design pour le concepteur des systèmes multimédias interactifs Faculté des sciences de l'éducation, Université Laval.
 Michard A. XML langaet applicat1999 Editions Eyrolles.
 Rohit Khare MCI Internet Architecture & Adam Rifkin California Institute of Technology. XML: a door to automated Web Applications, IEEE Internet Computing 1997.
 Martin P. La conception pédagogique de systèmes d'apprentissage multimédia interactif: fondements, méthodologie et problématique. Directeur du groupe de recherche GRAIM, Faculté des sciences de l'éducation, UniveréLava.
 Kannat S.E., Conil D., Sicard P. Une expérience de télé-enseignement pour autoroute de l'information LSR-IMAG.
 Java2 SDK standard edition. Available in http://java.sun.com/products/jdk/1.2