Proceedings of the 8th World Conference of the AIED Society, Kobe, Japan, 18-22 August 1997

Architecture of an Intelligent Tutoring System on the WWW

NTT Information and Communication Systems Laboratories
Musashino, Tokyo 180, Japan

NTT Information and Communication Systems Laboratories
Musashino, Tokyo 180, Japan

Yoshimasa KOIKE
Multimedia Communication Laboratories
NTT America, Inc. Pacific & Western Division
Mtn View, CA94040, U.S.
Phone:+1 415 940 6517
Fax:+1 415 940 1375

Yasuhisa KATO
NTT Information and Communication Systems Laboratories
Musashino, Tokyo 180, Japan

Hirofumi TOUHEI
NTT Information and Communication Systems Laboratories
Musashino, Tokyo 180, Japan

NTT Information and Communication Systems Laboratories
Musashino, Tokyo 180, Japan

Abstract : This paper presents the architecture of a WWW-based intelligent tutoring system (ITS) called CALAT. Using a conventional WWW browser on the client, students access CALAT server which provides an individual adaptation capability. The ITS kernel on the CALAT server, employing overlay model, presents the courseware pages so that the student can achieve a learning goal consisting of hierarchical sub-goals. Three types of pages are available in the CALAT courseware; explanation, exercise and simulation. Explanation page, presenting a material corresponding to the conceptual learning sub-goals, can be any type of HTML data including plain text, image, audio, JAVA applet, and/or plug-in application. It is also possible to use the HTML data residing on another WWW server as the explanation page. Exercise page is dynamically generated HTML form with advisory voice message. Three types of questions, true/false, selection, and description, are available. Each question is associated to the leaning sub-goals, and the answers are analyzed to update the student model. Simulation page provides an interactive simulation environment for the student to get a procedural knowledge on a target(simulated) system. Simulation environment is implemented with the combination of a special animation program on the client and a state transition machine on the server. The state transition machine is responsible to control the behavior of the target system as well as to monitor the student's action over the network. The monitored result is also used to update the student model. As the CALAT system is designed to fully exploit the WWW standards, its architecture is rather modular. Taking advantage of this modular architecture, further extension of CALAT is also discussed toward component-based ITS.

1 Introduction

It becomes very common to use the World-Wide Web(WWW) [1] system on the Internet or an intranet not just as a multimedia hypertext but as the infrastructure of the networked information system. In such a circumstances it is natural to apply the WWW for the educational or training purpose. With the WWW as an educational platform, it will be feasible for the students to access the multimedia courseware with general-purpose browsers. No special tools are required to start learning. For the courseware provider, it is not necessary to worry about the distribution and maintenance of the copies of the courseware but they just take care of the original on their server.

The WWW, however, provides only a passive and static hypertext with no individual adaptation capability. To offer not a static textbook but an interactive educational environment, it is necessary to extend the WWW framework so that it will be capable guide the students taking into account of their comprehension ability, area of interest, etc.

One possible solution toward this direction is to download the ITS courseware via the WWW. In this case all the ITS logic runs on the client side so it would be difficult to monitor the progress of each student. Other possibility is to implement the individual adaptation logic on the WWW server. There are several proposals in this line. Some of them have simple HTML modification logic [2] or interactive animation program [3]. Some equipped with the complete ITS logic on the server [4].

We have proposed CALAT [5,6], a WWW-based intelligent tutoring system(ITS). The system features a WWW server consisting of a tutoring system called CAIRNEY [7,8] which has been developed as a standalone ITS with easy-to-use authoring system. The user identification mechanism devised in the CALAT system makes the individual adaptation capability of the ITS available from the client with standard WWW browser and protocol.

After several improvement and streamlining process, several CALAT servers are currently in operation. This paper describes the architecture and implementation of the latest version of CALAT. Three types of pages are available in the CALAT courseware; explanation, exercise and simulation, each of which are associated with the learning sub-goals. Explanation page can be any type of HTML data including JAVA applet and/or multimedia plug-in application. It is possible to use the HTML data residing on another WWW server. Exercise page is dynamically generated HTML form used to analyzed the student's comprehension status. Simulation page allow the student to acquire a procedural knowledge on a target(simulated) system which is presented as a state transition machine.

As the above mentioned mechanism of the CALAT system is designed to fully exploit the WWW standards, its architecture is rather modular. Potential extension taking advantage of this modular architecture is also discussed toward component-based ITS which is attracting common interests at present [9].

In the next section, the architecture of CALAT is outlined. Then the latest implementation is described in detail. Finally the future extension is discussed.

2 System Overview

CALAT architecture is an extension of conventional WWW system shown in Figure 1. It is designed to provide individual adaptation capability while fully utilize the existing WWW protocols and browsers[5].

Figure 1. CALAT Architecture

The CALAT server consists of a WWW daemon process and back-end ITS processes. One ITS process is invoked to take care of one student. An user identification mechanism is introduced to maintain the correspondence between the ITS process and the student over the stateless protocol of the WWW. When the CALAT command button definition is sent from the server, the mechanism make every URL links representing the command include the student's account name. When the student click one of these URL links, the mechanism forwards the associated command to the ITS process responsible for the student by examining the account name in the URL.

CALAT ITS process implements the pedagogical logic featuring individual adaptation capability which makes it possible to present courseware pages by taking into account of the student's state of understanding dynamically. To achieve this capability, the ITS process consists of overlay student model and tutoring expert system. It deals with the courseware having tree-structured learning goals and pages associated to these goals(figure 2). There are three types of courseware pages; explanation, exercise, and simulation. According to the student's level of comprehension and the acquired learning goals, which are checked through the exercise and simulation, the ITS process determines the next display sequence of the explanation pages. For example, if the exercise result is so good that the process judges student's comprehension level is high, it teaches the student with the pages of brief explanation on the next learning goal. As another example, if the process judges that the mandatory learning goal is not acquired, it goes into the remedial phase with the detailed explanation pages on the mandatory learning goal.

CALAT courseware is generated with a special authoring subsystem. With the authoring system, courseware can be designed by simply defining the dependency between learning goals and the association between learning goals and the pages of explanation, exercise, and simulation.

Figure 2. ITS process and courseware pages

3 Implementation Detail

Three type of courseware pages, explanation, exercise, and simulation discussed in the previous section, are implemented with the specific mechanism which characterizes CALAT as an ITS on the WWW. These mechanisms are described in the following subsection.

3.1 Explanation Page

Explanation page presents a material describing or explaining the conceptual learning sub-goals of the courseware. As an explanation page, CALAT accepts any type of HTML data including plain text, image, audio, JAVA applet, and/or plug-in application such as VRML and Shockwave. Actually it is possible to use the HTML data pointed by URL as the explanation page. It means that any HTML data residing on the other WWW server can be included as a part of courseware.

The mechanism is shown in figure 3. Courseware sub-goal is associated with the URL pointing to the explanation page possibly on the other WWW server than CALAT.

When the student send a request, CALAT server determines the sub-goal to be explained according to the pedagogical logic, then responds with the JAVA script defining the CALAT command button as well as the procedure to access the explanation page of associated URL.

The WWW browser on the client will execute the procedure to display the explanation page. Figure 4 indicates the example of the explanation pages which is the part of a courseware on the U.S. presidential election.

Figure 3. Explanation page on another WWW server

Figure 4. Sample screen

3.2 Exercise

Exercise page is provided to examine the student's status of comprehension. Three types of questions, true/false, selection, and description, can be used. Exercise page is a HTML fill-out form dynamically generated from an exercise template (figure 5). Each question in the exercise template is associated to the leaning sub-goals.

When the student submits the form including wrong answer, ITS process updates the student model, then judges to give the student another chance to challenge the exercise or to guide him into the remedial phase. To let the student try the exercise again, the form is presented with appropriate advisory voice message. ITS process generates this form by modifying the exercise template to reflect the sub-goals acquired by the student.

Figure 5. Exercise

3.3 Interactive Simulation

Interactive simulation is an important facility to let the student acquire procedural knowledge. In the case of CALAT, interactive simulation page is associated with one of the learning sub-goals.

The simulation page consists of a state transition machine(STM) running on the server and a special animation program on the client[6]. (figure 6). The STM represents the behavior of the simulated(or target) system. The STM is defined as a collection of states, the input event which triggers the state transition, and the output actions taking place when certain state transition occurs. The input event and output action of the STM are linked to the animation program on the client.

The STM is also responsible to monitor the student's action. Each input event by the student's action is compared with the allowable action. When the final state is reached, the result is reported to the ITS process to judge if the student accomplish the learning goal associated with the simulation page.

The animation program on the client runs as a external viewer or a plug-in program of the WWW browser. It supports GUI(graphical user interface) objects(widgets) such as buttons or key-input fields, and multimedia outputs such as animated images or sounds. The student's click or type on these widgets causes the event input to the STM, and the output actions from the STM change the color of the widgets, redraw the window image or trigger sound output. All the communication between the STM and the animation program is implemented using standard HTTP. The student's action is sent with HTTP GET method with query string. The response from the STM representing the action of the target system is sent as a multimedia data with suitable mime-type.

This implementation benefits to make the simulation environment easily available on the multiple client platform, since just a GUI part without STM should be ported. Though our experience shows that the enough response speed is obtained with the standard T1 connection, the performance might suffer with the low speed network. To provide sufficient performance in such a circumstance, the STM should be implemented on the client side. Cross platform language such as Java will be useful for this purpose.

Figure 6. Interactive simulation

4 Future Extension

The CALAT architecture presented in the previous section could be extended in several ways toward component-based ITS.

One possibility is to implement more sophisticated interactive environment based on the current exercise framework. Recently there appears a lot of interactive multimedia tools which can be used as a plug-in application of the WWW browser. It is attractive and not very difficult to utilize the interactive animation developed with these tools instead of current fill-out form for the exercise page. Important point is that interface to associate learning sub-goals to the exercise page is already defined. The interactive animation can send the student's interaction results to the server using the same protocol used by the HTML fill-out form. It is also possible for the ITS process to control the behavior of the animation using the currently defined exercise template modification scheme. No modification is required to the ITS process or the structure of learning goals. The interactive animation defined in this way can be considered as an active ITS component associated with certain educational target which is reusable in the other courseware.

Another possibility is an adaptive hypertext mechanism as a extension of the explanation page. Current implementation of the explanation page does not allow the contents or the hypertext links in the page. By modifying the links to be directed to the CALAT server, it would be possible to monitor the students behavior and to modify the hypertext space to reflect the student's learning status[10]. These courseware pages associated with the learning goals can be also considered as the reusable educational resources distributed over the WWW system.

5 Conclusion

The architecture of a WWW-based ITS called CALAT is presented. In the current implementation, three types of pages, explanation, exercise and simulation are available in the courseware. They are implemented by fully utilizing the WWW standards. Further extension of CALAT is also discussed toward component-based ITS.

The current system has been already introduced on the intranet of our company for the employee training program. There are several coursewares for the novice level training in the field of telecommunication technology such as ISDN or local area network. CALAT server with several coursewares are also available on the Internet at


[1] Berners-Lee, T., et. al. (1994). The World-Wide Web. Communications of the ACM, 37(8), 76-82.
[2] Kay, J. and Kummerfeld, R.J., (1994). An Individualized course for the C programming language. Proceedings of the Second WWW Conference.
[3] Ibrahim, B. and Franklin, S.D. (1995). Advanced Educational Uses of the World-Wide Web, Proceedings of the Third International World Wide Web Conference '95: Technology, Tools and Applications.
[4] Schwarz, E., Brusilovsky, P., and Weber, G. (1996) World-wide intelligent textbooks. Proceedings of the the ED-MEDIA 96.
[5] Nakabayashi, K., Koike, Y., Maruyama, M., Touhei, H., Ishiuchi, S., and Fukuhara, Y. (1995). A distributed Intelligent-CAI System on the World-Wide Web. Proceedings of the ICCE 95, 214-221.
[6] Nakabayashi, Maruyama, M., K., Koike, Y., Fukuhara, Y., and Nakamura, Y., (1996). An Intelligent Tutoring System on the WWW Supporting Interactive Simulation Environment with a Multimedia Viewer Control Mechanism. Proceedings of the WebNet 96.
[7] Fukuhara, Y., Kimura, F., Kohama, C., and Nakamura, Y. (1995). A Knowledge-based Educational Environment Integrating Conceptual Knowledge and Procedural Knowledge in Telecommunication Service Field. Proceedings of the ED-MEDIA 95, 229-234.
[8] Fukuhara, Y., and Kiyama, M. (1993). A Knowledge-based Educational Environment for Real Business Settings. Proceedings of the World Conference on Artificial Intelligence in Education.
[9] P1484 Working and Study Groups,
[10] Maruyama, M., Nakabayashi, K., Koike, Y., Fukuhara, Y. and Nakamura, Y.(1996). An Approach to Implement Adaptive Hypermedia for an Intelligent Tutoring System on the World-Wide Web. Proceedings of the ED-MEDIA 96, 373.