Natalie Stash INTELLIGENT AGENTS FOR DISTANT LEARNING SYSTEMS
The paper discusses the way and privileges of using intelligent program agents (IPA) for developing distant learning systems (DLS).
IPA present a new approach in developing subsystems as parts of intelligent systems.
For developing the majority of modern distributed information systems in the Internet the architecture "client-server" is used. The most part of existing DLS advises students learning material only for passive examination or may possess test subsystem but its results do not have influence upon the structure of the learning material. For developing such systems service WWW can be used, besides, the most widespread is the use of expansion module on the basis of interface CGI or its subtypes: WinCGI, FastCGI, ISAPI NSAPI and so on. Combined with interfaces of access to database-management system expansion mechanisms of HTTP servers allow to fulfill tasks of keeping and processing any information about the student on the server. But this method is not good for developing adaptive DLS because complex processing of test results, building "intelligent" modules (working on the basis of knowledge base) loads server system greatly. Also data transferred from server to client contains much redundant information. The effectiveness of adaptive DLS can be achieved by transition part of computing from server to client. With this aim intelligent program agents are used. The most part of existing program agents can be classified with the following scheme (intelligent systems):
IPA in adaptive DLS are stationary learning hybrid, functioning in the framework "client-server". IPA is able to receive all necessary information from the DLS server and then develop learning plan adapted to the concrete student on the basis of knowledge base. Privileges of adaptive DLS using IPA as compared with the systems based on expansion modules of HTTP server are the following:
-
Loading on the DLS server resources is not much. Necessary computational recourses taking into consideration one user are comparative with usual loading on the WWW server without DLS.
-
During learning process in the main learning materials are transferred through the net. IPA is loaded either once during the whole learning period or before each work session with the adaptive DLS server (in dependence on the chosen architecture of the system). Student model loading into the memory of intelligent agent - single transaction in the beginning of work session with adaptive DLS. If it is necessary IPA is able to address to DLS server for all necessary information, for example, for test results.
-
Combination of the first and second privileges results in high scaling of DLS. Loading on the server during adding new clients increases much slower as compared to the systems based on the expansion modules of HTTP server.
But while developing adaptive DLS using IPA it is necessary to solve several problems, which other systems do not have.
-
IPA can have number of properties defining its inner state. Actually student model is a subset of IPA states. It is necessary to keep agent state between work sessions with DLS.
-
Taking into consideration heterogeneous character of computational systems functioning in the Internet, IPA must be able to function on different platforms in various environments (operational systems). Besides, functioning in the structure of WWW browser IPA must be compatible with the main applied in the net browsers.
The system can be realized in such the way that IPA itself carries out all the operations with the learning materials - graphical inference, sound accompaniment and so on. In such the case it must be independent supplement of the operational system. But while developing such IPA it is necessary to solve the problem of displaying various forms of information (text, image, audio, video and so on) that makes process of IPA realization rather durable and demanding for resources. At the same time existing browsers are able to display a great number of information forms. So the best way of IPA realization is the method when agent while collaborating with browser uses it for displaying all the learning materials and only navigates the student through the learning material in conformity with the rules.
The work is partially supported by Russian Foundation for Basic Research (grand 98-01-0081).
References
-
Gavrilova T., Voinov A. (1995) Adaptive Interface Design and Scenario Control via User Modeling. In: Preprints of IFAC/IFIP/IFORS/IEA Simposium on Analysis, Design and evaluation of Man-Machine Systems, MMS'95. MIT, Cambridge, 535-540.
-
Gavrilova T., Voinov A., Zudilova E. (1995) User Modeling Technology in intelligent system design and interaction. In: Proceedings of East-West International Conference on Human-Computer Interaction HCI'95, Moscow, 115-128.
-
Gavrilova T., Averbukh E., Voinov A. (1996) An Approach to Learning Software Based on Student Modeling. Lecture Notes in Computer Science N 1108: Third International Conference on Computer Aided Learning and Instruction in Science and Engineering. CALISCE'96, San-Sebastian, 448-450.
-
Chernigovskaya T., Gavrilova T., Voinov A., Udaltsov S. (1998) Intelligent Development Tool for Adaptive Courseware on WWW. Proc. of 4th Int. Conf. On Comp. Aided Learning and Instruction in Sc. And Eng., 464-467.
-
Kaas R. (1989) Student Modelling in Intelligent Tutoring systems - Implications for User Modeling. In: User Models in Dialog Systems. USA.
L.V.Kotova, N.A. Yarmosh TESTING IN COMPUTER-ASSISTED LEARNING SYSTEMS
The features of the authoring tools for tutoring software creation are considered in this paper. Nowadays the growth of interest to the authoring tools in the domain of Computer-assisted learning (CAL) is stipulated by a number of reasons, among which the acceleration of creation process of the system and an openness of the project giving a users the possibility to modify a computer-assisted learning system, the user experience in programming being minor.
The following figure illustrates the main advantages of CAL-system:
Such an authoring tools for creation of the CAL-system AOSProject is developed in the Institute of Engineering Cybernetics of the National Academy of Sciences of Belarus [1]. This system is universal and can be filled with a material on any discipline. The given software allows teachers to create a tutorial system and tests. The development of the new version of the CAL-system named AOSControl to check knowledge under the text and graphics answers is now carried on.
The assessment with use of the following text and graphics questions is realized in this system: selective answers, freely constructed answers, freely formulated answers, ordering of the answers.
The test requiring the selective answer consists of the problem text possibly with an illustration and texts or illustrations of answer versions. It is necessary to mark correct variants of the answers. This group ensures knowledge testing at a level of identification, distinction and classification. As for the freely constructed answer fragments of sentences or graphics objects is offered for learner. In this case, it is necessary to compile the answer with the help of these fragments. The form demanding freely formulated answer contains the text of problem and input area. Learner enters the answer text using the keyboard or letter keys located in the alphabetic order on the page. The tests of this group allow to check learner abilities to situations description from the given data domain. Author of learning course inputs the information about possible ways of the description for each situation or creates reliable tools of the automatic analysis of the entry text and addition of a knowledge base. The test requiring the ordering of answers consists of problem statement and texts or illustration of possible variant of the answers that are necessary for ordering on indicated criterions.
The unit of assessment forms a final learner's knowledge evaluation in numbers or percents taking into account following criterions: a correctness of task execution; complexity of the offered task (question weight); coefficient of importance of an investigated theme from which the task are offered; call to the help information; execution time of the task. The obtained evaluation is represented in the form "Testing results" and is fixed in the database about the learner.
References
1. L.V.Kotova, S.F.Lipnitsky, A.A.Protskevich, M.S.Shibut, N.A. Yarmosh. Creation of computer textbooks by tools of the system AOSProject // In: Proc. of Conf. devoted 65 years BSEU. — 1998. —BSEU, Minsk, Belarus. — Pp.82-83. (In Russian).
Share with your friends: |