AHyCo: a Web-Based Adaptive Hypermedia Courseware System

Journal of Computing and Information Technology - CIT 13, 2005, 3, 165–176 165

AHyCo: a Web-Based AdaptiveHypermedia Courseware System

Natasa Hoic-Bozic1 and Vedran Mornar21 Department of Computer Science, Faculty of Philosophy, University of Rijeka, Croatia2 Faculty of Electrical Engineering and Computing, University of Zagreb, Croatia

Adaptive hypermedia courseware systems resolve theproblem of users’ disorientation in hyperspace throughthe adaptive navigation and presentation support. We de-scribe the AHyCo �Adaptive Hypermedia Courseware�— an adaptive Web-based educational system for cre-ation and reuse of adaptive courseware with emphasis onadaptive navigation support and lessons sequencing. Theproposedmodel consists of the domainmodel, the studentmodel, and the adaptive model. The system is composedof two environments: the authoring environment and thelearning environment.

Keywords: Web-based adaptive educational systems,adaptive hypermedia, adaptive navigation support,courseware authoring, AHyCo �Adaptive HypermediaCourseware� system.

1. Introduction

Traditional computer-aided education has beengreatly enhanced by utilizing the Web-based hy-permedia systems. Despite their advantages,some problems related to the usage of suchWWW systems become apparent as well, par-ticularly concerning the disorientation of theusers, or the “getting lost in hyperspace” prob-lem �Maurer & Scherbakov 1996�. An adaptivehypermedia system �AHS� resolves this prob-lem by adapting the presentation of hyperme-dia content or links, based on the user model�Brusilovsky 1996�.

In this paperwedescribe theAHyCo—anadap-tive educational hypermedia system �AEHS�for creation and reuse of adaptive coursewarewith emphasis on adaptive navigation supportand lessons sequencing. The purpose of lessonsor curriculum sequencing technology is to pro-vide a student with the most suitable sequence

of knowledge units to learn �Hubscher 2000�.It helps the student to find an optimal naviga-tional path through the material to be learned�Hoic-Bozic & Mornar 2001a�.

The model we propose marks all the links andsuggests which page the student should visit,according to the student’s knowledge and somespecific attributes of each lesson. A main prin-ciple for our approach is to maximize the spacethat the user may explore and to accomplish atrade-off between free exploration and guidedexploration.

The proposed model consists of the domainmodel, the student model, and the adaptivemodel. The system is composed of two en-vironments: the authoring environment and thelearning environment.

The remainder of this paper is organized as fol-lows: in section II some more recent achieve-ments in the area of adaptive hypermedia sys-tems have been mentioned, section III describesthe domain model, student model and adapta-tion model of the AHyCo system, section IVdescribes some of the implementation issues,section V presents the authoring and learningenvironments, section VI describes the use andevaluation of the AHyCo system, section VIIpresents conclusions and future plans.

2. Background

According to Brusilovsky �1996�, with the termadaptive hypermedia systems we denote all hy-pertext and hypermedia systems that reflect some

166 AHyCo: a Web-Based Adaptive Hypermedia Courseware System

features of the user in the user model and ap-ply this model to adapt various visible aspectsof the system to the user. An adaptive hyper-media system �AHS� adapts the presentation ofcontent or links, based on the user model. Twomajor technologies in adaptive hypermedia aredistinguished: adaptive presentation and adap-tive navigation support. Adaptive presentationadapts either the content of a document or thestyle of the text. Adaptive navigation supportconcentrates on changing the presentation oflinks.

The most popular area for adaptive hypermediaresearch is the educational hypermedia, wherethe goal of a student is to learn the material ona particular subject �Brusilovsky, 1996�. Themost important element in educational hyper-media is the user knowledge of the subject thatis being taught. Certain student may know al-most nothing about the same lesson that maybe trivial and boring for another. In both casesthe students need navigational help to find theirway through the knowledge space.

A number of first generation adaptive hyperme-dia systems �Carver, Hill, & Pooch, 1999�werebuilt between 1985 and 1993. They were gener-ally standalone PC or Macintosh-based systemswith limited adaptability through stereotype-based user models and limited adaptation tech-niques. ISIS-Tutor is a good example of afirst-generation adaptive system �Brusilovsky,& Pesin, 1994�.

Since 1993 the Web has become the primaryplatform for developing educationalAHS �Brusi-lovsky, 1999�. These second-generation AHSwere generally platform independent. They in-troduced new features such as adaptive multi-media. Some examples are ELM-ART �Brusi-lovsky, Schwarz, & Weber, 1996�, InterBook�Eklund, & Brusilovsky, 1998�, DCG �Vas-sileva, 1997�, AHM �Da Silva, 1998�, CALAT�Nakabayashi, 1997�, KBS Hyperbook �Henze,& Nejdl, 2000�, ALICE �Kavcic, 2001�, AHA�De Bra, & Ruiter, 2001� and AHA! �De Bra, etal. 2003�, NetCoach �Weber, et al. 2001�, ALE�Specht, et al., 2001�.

The second-generation AHS mostly use link an-notation. A variant of the overlay model forrepresenting the student’s knowledge is used,sometimes in combinationwith stereotypes. Theeducational state of the concepts from the do-main model is updated if user visits the page

that presents the concept’s content. Some ofthe AHS �CALAT, InterBook, ELM-ART, KBSHyperbook, ALICE, NetCoach� use tests as ad-ditional and more reliable criteria.

The most important shortcoming of an AHS isthe authoring part. The process of authoringshould include the development of the actualhypermedia content �lessons, tests, etc.� andthe definition of the rules for adaptation.

In order to motivate more authors to use theadaptive hypermedia, the authoring processshould be made much simpler than in someexisting GUI-based authoring tools �NetCoach,ALE�. The authoring component should enablethe straightforward creation of concepts, thelinkage of concepts by prerequisite relation-ships, and easy generation of the test questions.It should be user-friendly enough to enable aperson who is not a computer expert to designthe courseware. That includes the developmentof a graphic editor for concept networks, whichwill enable the authors to define the prerequisiterelationships with a drag-and-drop interface.

In our AHS AHyCo, particular attention is givento the authoring component of the system.AHyCo user interface is form-based. We strictlyseparate the learning dependencies �prerequi-site relationships� from the actual content �mul-timedia fragments�. This separation allows theauthors to use the same set of fragments froma domain to build diverse courseware by sim-ply defining new prerequisite relationships andthe values for adaptation rules �Hoic-Bozic, &Mornar, 2003�.

AHyCo attempts to provide a complete course-ware management system for practical courses.AHyCo is delivered as an Open Source software�AHyCo, 2003�. It uses an easy to use graph-ical drag-and-drop interface for the definitionof prerequisite relations between learning ob-jects. AHyCo is designed to enable the authorsto develop adaptive learning courses without theknowledge of programming.

3. The Model of the AHyCo

Our model of an adaptive educational hyper-media system consists of the domain model,which describes the structure of the learning

AHyCo: a Web-Based Adaptive Hypermedia Courseware System 167

domain as a set of reusable concepts linked to-gether with prerequisite relationships, the stu-dent model encompassing the student’s know-ledge of the learning concepts, and the adap-tive model which contains rules for adaptation�Hoic-Bozic & Mornar, 2001b�.

3.1. Domain Model

A domain model of the proposed AHS has atwo-level structure and consists of concepts aselementary pieces of knowledge for the givenlearning domain �Fig. 1�.

For the first domain level the AHS use the formof a graph �Ck� LCk�, where Ck is the set of con-cepts and LCk is the set of arcs, LCk � Ck �Ck.Links represent the prerequisite relationships� that denote pedagogical constraints, for ex-ample Ci � Cj means “concept Ci should belearned before concept Cj”. The AHS distin-guish between the lessons Ci and tests Tj asthe concepts or graph nodes. Tests contain thequestions about the domain lessons.

A lesson Ci is defined as a

�FCi� PCi�Qi� Ri� wci� lMyi�

where:

FCi is a set of multimedia fragments �smallbuilding blocks, e.g. a piece of text, graphics,sound, video clip� � � �.

PCi is a set of prerequisite concepts, whichare essential for the student to understand thelesson Ci.

Qi is a set of questions related to lesson Ci.All questions are multiple-choice�multiple-or-single-answer, so each question is definedas �FQ� q�A� a� B� where:

FQ is the set of multimedia fragments thatform the question.

q is the weight of the question or the con-fidence level of the fact that student ei-ther knows the lesson if he�she answersthe question correctly, or does not knowthe lesson if he�she answers incorrectly,q � �0� 1�.

A is the set of offered answers. Each of-fered answer Aj is an ordered pair. Thefirst element is a set of multimedia frag-ments, or a function fj�p1� p2� � � � � pn� thatevaluates a candidate’s answer on the basisof parameters p1, p2,� � � , pn. Function fjis defined in a scripting language and willbe evaluated after the random generationof the parameters. The second elementis vi, the value of the ith answer, whichcan be positive �for correct answers� ornegative. More obvious answers generallyshould have a smaller value.

B is the set of lower and upper bounds forparameters p1, p2,� � � , pn.

Ri is the rank of the lesson Ci calculated as

R0 � 0Rk � maxRj � 1�

j j Cj � PCk�

Fig. 1. An example of a domain model.


wci is the weight of the lesson, wci � �0� 1�.

lMyi is the minimum acceptable knowledgelevel for MYCIN model, lMyi � ��1� 1�.

A test Tj is defined as a �PTj� nj�Nj� Rj� where:

PTj is a set of prerequisite concepts-lessons,PTj � fCi j Ci � Tjg. Tj will contain thequestions related to lessons Ci.

nj is the total number of questions in Tj.

Nj is the set of configuration rules that spec-ify how many questions for each conceptCi � PTj are placed into the test Tj.

Rj is the rank of the test Tj.

To split the domain into more manageable units,concepts are grouped into modules Mk. Thesecond level of the domain model is a directedgraph D � �M� LM�, where M is the set ofmodules and LM is the set of arcs, LM �M�M. The arc connecting modules Mk andMl exists if Mk � Ml. The directed graph Drepresents the class the student has enrolled.

A moduleMk is defined as a �Ck� PMk� lmk� Rk�where:

Ck is a set of concepts that create the module.

PMk is a set of prerequisite modules formodule Mk.

lmk is the minimum acceptable knowledgefor module, lmk � ��1� 1�.

Rk is the rank of the module Mk.

The domain model distinguishes between thetwo kinds of tests: the mini-tests or quizzes Tkfor modification of the navigation within themodule, and one final test Tf for the navigationbetween the modules. The final test is the nodewith the highest rank in the graph �Ck� LCk�.

The proposed representation for the domainstructure is suitable for storing learning ma-terials from different areas �computer science,mathematics, medicine, art, etc.�. The struc-ture of the knowledge is not necessarily hi-erarchical �chapters, subchapters, pages� butrather concept-oriented �Hoic-Bozic & Mornar,2001a�.

The pedagogical constraints between the lessonsdepend on the subject area. For example, formath lessons, Ci � Cj usually means that it isimpossible for a student to start learning Cj if

he�she has not gained the knowledge of Ci. Forsome other subject areas, the prerequisite rela-tionships simply denote the sequence of learn-ing, proposed by the author. For the same setof lessons, another teacher may choose a com-pletely different learning sequence, i.e. prereq-uisite relationships.

3.2. Student Model

For the AHyCo system, a two-level studentmodel as a variant of the overlay model �Brusi-lovsky, 1996� for representing the student’sknowledge is proposed. The first level repre-sents the estimate of students’ knowledge aboutthe lessons. The second level represents theknowledge about the modules. For every lessonCi the main attributes recorded for each studentare ri and ki. For every Mk the AHS is recordedthe knowledge value about module kmk.

ri is the estimate about whether the student hasread the lesson Ci or not. ki is the estimateabout the student’s knowledge of the lesson Ciand is calculated using a variant of the MYCINmodel, a widely used expert systems’ model�Anjaneyulu 1997, Ng & Abramson, 1990�.

The knowledge value of a lesson is set by test-ing and can range from �1 �student does notknow the lesson� to 1 �student knows the les-son�. Before the student takes any of the tests,all lessons in the student model have an initialvalue of ki � 0.

After answering each test question related tolesson Ci, the new knowledge value k�i for thelesson Ci is calculated according to �1�. Thenew value k�i is based on the previous knowledgevalue ki and the factor q, the question weightor the confidence level of the fact that the stu-dent knows or does not know the lesson. If thestudent answers the question correctly, f � q,otherwise f � �q.

k�

i �

��

ki � �1 � ki� � f� ki � 0� f � 0ki � �1 � ki� � f� ki � 0� f � 0 �1��ki � f ��1 � min�jkij� j f j�� otherwise

The model asymptotically increases�decreasesthe knowledge valuewith each correct�incorrectanswer according to the previous knowledgevalue ki and the question weight q from thedomain model �Hoic-Bozic & Mornar, 2001b�.


The knowledge value kmk of a module Mk is setaccording to �2�.

kmk �X

jjC1�Ck

kj � wcj �2�

According to the formula, more important les-sons for the module �with higher weight wcj�have greater influence when calculating theknowledge level kmk.

3.3. Adaptation Model

The adaptation model consists of adaptationrules that define how the domain model and thestudentmodel are combined to perform adaptivenavigation support.

In our system, we employed the adaptive nav-igation, which is a combination of free �open�and guided �forced� navigation. The studentcan freely follow any hyperlink within a mod-ule or graph �Ck� LCk�, but a list of hyperlinks isoffered that suit him best, according to the navi-gation plan generated for him. The AHyCo usethe combination of link sorting and link annota-tion adaptive techniques. The navigation withina graph �M� LM� is restricted and depends onthe student’s knowledge value kmk.

According to the student model and currentlydisplayed lesson Ca, the concepts from moduleMk are classified into several subsets: learnedconcepts CLa, recommended concepts whereall prerequisite concepts have been visited CCa,and not recommended concepts CNa. There arealso completely recommended concepts CPa orrecommended concepts that are in direct pre-requisite relationship with Ca.

Concepts Ci from the Ck of the module Mk areclassified according to the algorithm 1:

Input: graph �Ck� LCk�, active lesson Ca � Ck

Output: sets CLa, CPa, CCa, CNa

CLa � �, CPa � �, CCa � �, CNa � �for each Ci � CknfCag

if ri � true�* Ci is visited *�CLa � CLa � fCig

else if Ca � Ciif rj � true� �Cj � PCinfCag

�* all prerequisites for Ci are

visited except Ca –Ci is completely recommended *�CPa � CPa � fCig

else�* Ci is not recommended *�CNa � CNa � fCig

else if rj � true� �Cj � PCi�* all prerequisites for Ci are visited– Ci is recommended *�CCa � CCa � fCig

else�* Ci is not recommended *�CNa � CNa � fCig

Algorithm 1. Classifying the concepts Ci � Ck.

Navigation within the module Mk goes on be-fore the student solves the final test Tf . Thisnavigation is actually the traversing of a directedgraph �Ck� LCk�, following the hyperlinks sug-gested by the system on the bottom of the page.The model uses mini-tests or quizzes Tj to checkthe students’ knowledge and to correct the stu-dent model while he�she navigates within themodule. Transition to another module is possi-ble after the successful completion of the finaltest Tf �Hoic-Bozic & Mornar, 2001a�. Thereare three possible outcomes of the test Tf be-longing to the module Mk:

Tf is completely passed — the Mk is learned:knowledge value kmk � lmk and ki � lMyi,�Ci � Ck. The student can proceed to an-other module according to the directed graph�M� LM�;

Tf is partially passed — the concepts Ci withki lMyi are offered for repetition but thestudent can proceed to another module sincethe kmk � lmk; Mk is partially learned;

Tk is not passed — kmk lmk the conceptsCi with ki lMyi from Mk are offered forrepetition and the student should retake thetestTf in order to proceed to anothermodule.

The student model is updated after the test Tjhas been solved according to the algorithm 2:

Input: the studentmodel, lmk and lMyi from thedomain model, results of the test Tj, Ci � PTj

Output: updated values of ki, rpi, kmi


use MYCIN for calculating kiif Ci � PTj, ki lMyi

set rpi � 1Mk is partially learned

if Tj � Tf�* Tj is final test for Mk *�calculate value kmk of Mkif kmk � lmk

Mk is learnedelse Mk is partially learned

Algorithm 2. Updating the student model after Tj.

4. Implementation Issues

In the development of the AHyCo system werely on the Microsoft .NET technology. ASP.NET provides a new server-side control ar-chitecture which facilitates the development ofhighly interactive web pages. It has an event-based programming model, making web devel-opment much more like traditional VB formsprogramming. An average ASP.NET page re-quires less code than an equivalent ASP page,which leads to greater developer productivityand better maintainability. ASP.NET pages arealso compiled, soweb servers runningASP.NETapplications significantly exceed the perfor-mance and scalability levels of previous ASPapplications �Anderson et al. 2001�.

The AHyCo network application is based on therelational database management system. Thesystem components �Fig. 2� are:

a� relational database

b� authoring interface

c� midle-tier component for communication be-tween Web application and database

d� ASP.NET Web application.

4.1. Relational Database

All information about the subject matter and thestudents are stored in a Microsoft SQL Server2000 database.

Themain part of the databasemodel is the learn-ing and testing subschema. The formulas andsmaller multimedia fragments of lessons andquestions are stored in the database as binary ob-jects. Larger multimedia objects, such as videoor audio clips, are stored in the file system, andare connected to the lessons by hyperlinks.

For security reasons, the set of IP addresses thatare allowed to access the system can be speci-fied.

4.2. Authoring Interface

AHyCo system uses Microsoft Access forms asan interface for the authors. Because the lessonsand the questions are stored in the database asWord objects, the author can use the familiarMicrosoft Word application for updating. Word

Fig. 2. The components of the AHyCo network application.


is chosen because of its programming features:the fragments FCi can be easily connected, up-dated and converted to HTML or PDF format.To enhance performance of the system and en-sure better scalability, the lessons and questionsare also stored on the Web server as sets ofHTMLdocuments during the courseware prepa-ration phase.

The author names the lessons Ci and groupsthem intomodulesMk. Themodules are groupedinto subjects. The prerequisite relationshipsbeetween concepts and modules are defined.

The author should also determine the weightof the questions �q� and lessons �wcj� and theminimum acceptable knowledge level for eachconcept in the module �lMyi� and for modulesthemselves �lmk�. This elements form the vari-able part of the adaptation rules.

In our system we separate the learning de-pendencies �prerequisite relationships� and thevariable part of the adaptation rules from theactual content �multimedia fragments�. Thisseparation allows the authors to use the sameset of fragments from a domain to build diversecourseware by simply defining new prerequisiterelationships and values for the adaptation rules�Hoic-Bozic & Mornar 2001a�.

4.3. Middle-tier Component

The midle-tier component is responsible forcommunication between the Web applicationand the database. It contains the logic of thesystem — the rules for adaptation �Anderson etal. 2001�.

To generate the pages, the system consults themiddle-tier component. This component deter-mines which content will be shown next, ac-cording to the adaptation rules and informa-tion from the domain model and the studentmodel. Based on the information returned fromthe middle-tier component, appropriate HTMLfragments are composed together by the Webapplication that is responsible for displaying thelesson or test question simultaneously with therest of the page �hyperlinks and buttons for nav-igation�.

4.4. Web Application

The learning environment is implemented as aMicrosoft ASP.NET C# web application. Thisis the only part of the system with which thestudents interact. They use the Web browser-based interface to the learning environment ofthe AHyCo courseware.

The web application needs to perform the fol-lowing tasks: login and validate the student, listsubjects related to the student, display lessons’content and navigational elements, display tests’questions and submit the students’ answers, dis-play the test results.

Web pages presented to a student are gener-ated adaptively �on the fly� when the studentrequests them, based on the contents extractedfrom the database. There are three kinds ofpages: lessons, questions and special pages�e.g. login page, help, and test results page�.

5. Authoring and Learning Environments

The system is composed of two environments:the authoring environment and the learning en-vironment. The authoring environment is usedby teachers to define adaptive courseware ma-terials in various learning domains. The web-based learning environment allows the studentto log in and study automatically generatedcourseware, which dynamically adapts accord-ing to his�her success in acquiring knowledge.

5.1. Authoring

Authoring is the crucial task in adaptive hy-permedia design. It includes the developmentof both the network of lessons and the tests.Before the courseware construction has started,the author should divide the subject matter intosmaller parts �concepts or lessons�. The lessonsshould be connected by prerequisite relationsand grouped into modules.

The first authoring step includes creation of theset of hypermedia fragments FCi �text, image,� � � � that represent the content of the lesson Ci�Fig. 3�. The lessons are stored in the databaseas Word, Excel or PowerPoint objects, so theauthor can use the familiar Microsoft Office ap-plications for preparation and updating.


Fig. 3. MS Access form with Word OLE object that contains the text of a lesson.

During the authoring phase, the author sets therules for adaptation by defining the prerequi-site graphs of concepts �C� LC� and modules�M� LM�. The graphs are created and con-nected by using a graphic editor with a drag-and-drop user interface �Fig 4.�.

When defining the lessons in a module, the au-thor determines the weight wci of the lesson forthe particular module, as well as the minimumacceptable knowledge level for each lesson inthe module lMyi. When defining the test ques-tions, the author defines the weight q of thequestions for the associated lesson. Tests will

be created randomly, so explicit enumeration ofquestions is not necessary, only the structure ofthe test should be specified. For each modulein a domain, the minimum knowledge level lmkis set. Hence non-technical educators are ableto use the adaptive features in a simple manner.The authors do not need to take into account anyother adaptation rules, or to use any other tool.The authoring process is thus concentrated onthe definition of prerequisites and on the estab-lishment of the difficulty level for lessons andtests. The responsibility for adaptive navigationremains on the AHyCo system �Hoic-Bozic &Mornar, 2001b�.

Fig. 4. Creating a graph of concepts.


5.2. Learning Environment

To use the AHyCo learning environment, a stu-dent has to log in.

After the process of authorization, the studenthas to choose the subject for learning. For theselected subjectD, the Web page containing thelesson Ci is generated. This lesson is chosen inaccordance with adaptation rules and the datastored in the student model, to correspond withthe students’ previous knowledge. The upperpart of the page �Fig. 5� is static and representsthe content of a lesson.

At the bottomof the page, hyperlinks to the con-tinuation lessons or tests Tj proposed by the sys-tem are enumerated. The suggested hyperlinksare automatically generated before the page isshown and are annotated with various colorscorresponding to concept types.

The concepts are listed in the following order:

� Completely recommended or main concepts— green color annotates the concepts whereall prerequisite concepts have been visited

and these concepts are the best continua-tion for Ci according to the directed graph�Ck� LCk�.

� Recommended concepts — orange color an-notates all other concepts where all prereq-uisite concepts have been read.

� Not recommended concepts – red color an-notates the concepts where some of the pre-requisite concepts have not been read orthe knowledge level of some prerequisiteki lMyi.

� Visited concepts — blue color annotates thelessons with ri � true or ki � lMyi.

All the hyperlinks within the hypertext networkare functional, so the student can follow any hy-perlink. The idea of free navigation is only tosupport and aid students. It is up to individualstudent if he�she will follow the system’s sug-gestions. After the student has finished learningof the module’s lessons, he�she should choosethe final test button. Test questions �Fig. 6� andthe sequence of the offered answers are gener-ated randomly.

Fig. 5. The page with the content of a lesson.


Fig. 6. The question page.

Transition to another module is possible afterthe successful completion of a test Tf . Out-comes of the tests by individual lessons are dis-played separately.

6. Using and Evaluating the AHyCoSystem

The AHyCo system is currently being used inteaching the students at the Faculty of Philoso-phy, University of Rijeka.

The purpose of this research was partly to ex-plore how the use of adaptive hypermedia sys-tem improves students’ learning and to find outthe students’ attitude concerning theAHyCo us-age.

However, the main goal of the AHyCo systemevaluation was to find out how the adaptationrules could be corrected according to the stu-dents’ results in gathering the knowledge. Thesystem enables the diagnostic of the success inthe adaptive courseware authoring. Accordingto the students’ knowledge results, the teacherfinds out possible shortcomings in the devel-oped learning materials. After that, the author

should correct the adaptation rules. In that way,the teachers – authors learn how to create andstructure learning materials for use withAHyCosystem in a better way.

A research has been conducted at the Depart-ment of Information Science, Faculty of Philos-ophy, Rijeka on 19 senior students of Informa-tion Science in the context of the class “Seminaron Teaching Methods in Information Science”.Part of the class’s subject matter has been pre-sented as six AHyCo modules. Each modulehas various number of concepts linked togetherwith prerequisite relationships.

In the context of the class, the students, as futureteachers in schools, learn how to use informa-tion and communication technology in educa-tion. So AHyCo system has been useful notonly as a learning tool, but also as an exampleof using computer technology and new teachingmethods.

The students learned how to use AHyCo sys-tem during the summer semester of the aca-demic year 2001�2002. They used computers inPC-room with LAN connected with modem


connection toCroatianAcademicResearchNet-work �CARNet� or to their own home comput-ers.

After the students had finished learning withAHyCo, the evaluation of the system was per-formed according to their knowledge levelsabout the concepts �ki� and modules �kmk�.

Despite of the fact that the main purpose of theevaluationwas not to show the level of students’success in learning, we could mention that allthe students accomplished the goal and success-fully solved the final test Tf of the last moduleM6.

Actually, the main goal was to find out how tocorrect the variable part of the adaptation rules�the question weight q, the lesson weight wci,the MYCIN level lMyi, the module level lmk�and to verify if the prerequisite relationshipswere correctly defined �prerequisite graphs ofmodules �M� LM� and concepts �C� LC��.

Analyzing the basic statistic data �mean, me-dian, standard deviation, quartile values� andthe correlation coefficients for ki and kmk, aswell as by cluster analysis, recommendationsfor the authors were proposed about the correc-tions of the adaptation rules.

In order to explore the students’ attitude con-cerning the AHyCo usage, the questionnaireabout the effectiveness and quality of AHyCoand the level of students’ acceptance of AHyCoas a teaching resource was developed. Accord-ing to the questionnaire results, the students ac-cepted the new way of learning with AHyCosystem. The main difficulty regarding the learn-ing conditions was in gaining access to AHyCocourseware since computers at the Faculty werenot available all the time.

7. Conclusion and Future Work

In this paper we have discussed the model andimplementation of AHyCo system for develop-ment and distribution of the adaptiveWeb-basedcourseware. The navigation model is based onadaptive lessons sequencing and adaptive navi-gation support techniques. To accomplish atrade-off between free and guided navigation,the list of suggested hyperlinks is offered at the

bottom of the page, in the optimal order gen-erated by the system. However, a student isallowed to follow any of them.

In order to verify the results, the sample course-ware was generated at the Department of Com-puter Science, Faculty of Philosophy, Rijeka.The main purpose of the evaluation was to findout how adaptation rules could be corrected ac-cording to the students’ knowledge levels aboutthe concepts and modules.

According to the results of statistical analysis,recommendations for authors have been pro-posed with regard to correction of adaptationrules and improvement of prerequisite graphs.

In the further development of this research,the developed AHyCo system will be improvedbased on the results obtained during the devel-opment and evaluation. We will modify thesample AHyCo courseware according to theproposed rules and verify the new version onthe next generation of students.

References

�1� AHyCo Home Page �in Croatian�, �2003�,http��AHyCo�zpm�fer�hr�

�2� R. ANDERSON ET AL. ASP.NET, Wrox Press, Birm-ingham, 2001.

�3� K. ANJANEYULU, Concept Level Modelling on theWWW”, Proceedings of the AI-ED 97 8th WorldConference on Artificial Intelligence in Education,Knowledge and Media in Learning Systems, �1997�,Kobe, Japan.

�4� P. BRUSILOVSKY, L. PESIN, ISIS-Tutor: an intelli-gent learning environment for CDS�ISIS users, Pro-ceedings of the interdisciplinary workshop on com-plex learning in computer environments �CLCE’94�,�1994�, Joensuu, Finland.

�5� P. BRUSILOVSKY, Methods and Techniques of Adap-tive Hypermedia, User Modeling and User-AdaptedInteraction, 6 �1996�, pp. 87–129.

�6� P. BRUSILOVSKY, E. SCHWARZ, G. WEBER, ELM-ART: an intelligent tutoring system on World WideWeb, Proceedings of the Third International Con-ference on Intelligent Tutoring Systems ITS’96,�1996�, Montreal, Canada.

�7� P. BRUSILOVSKY, Adaptive and intelligent technolo-gies for Web-based Education, Kunstliche Intel-ligenz, Special issue on intelligent systems andteleteaching, 4 �1999�, pp. 19–25.

�8� A.C. CARVER, J.M.D. HILL, U.W. POOCH, Thirdgeneration adaptive hypermedia systems, Proceed-ings of WebNet 99, World Conference on the WWWand Internet, �1999�, Honolulu, Hawaii.


�9� D.P.DA SILVA,Concepts and documents for adaptiveeducational hypermedia: a model and a prototype,Proceedings of the 2nd Workshop on Adaptive Hy-pertext and Hypermedia HYPERTEXT’98, �1998�,Pittsburgh, USA.

�10� P. DE BRA, J.P. RUITER, AHA! Adaptive hyperme-dia for all, Proceedings of the WebNet Conference,�2001�, Orlando, Florida.

�11� P. DE BRA ET AL. AHA! The Adaptive HypermediaArchitecture, Proceedings of the ACM HypertextConference, �2003�, Nottingham, UK.

�12� J. EKLUND, P. BRUSILOVSKY, Individualising in-teraction in Web-based instructional systems inhigher education, Presented at The Apple Univer-sity Consortium’s Academic Conference, �1998�,Melbourne, Australia.

�13� N. HENZE, W. NEJDL, Extendible adaptive hyper-media courseware: Integrating different courses andWeb material, Proceedings of Adaptive Hypermediaand Adaptive Web-Based Systems AH 2000, �2000�,Trento, Italy.

�14� N. HOIC-BOZIC, V. MORNAR, An Approach to Navi-gation Support inAdaptiveEducationalHypermediaSystem, Proceedings of Multimedia and Hyper-media Systems Conference MIPRO 2001, �2001�,Opatija, Croatia.

�15� N. HOIC-BOZIC, V. MORNAR, Navigation Supportin a Web-Based Adaptive Educational Hyperme-dia System, Proceedings of AMCIS 2001, �2001�,Boston, Massachusetts.

�16� N. HOIC-BOZIC, V. MORNAR, An approach to adap-tive hypermedia courseware authoring, Proceedingsof Hypermedia and Grid Systems, MIPRO 2003,�2003�, Opatija, Croatia.

�17� R. HUBSCHER, Logically Optimal Curriculum Se-quences for Adaptive Hypermedia Systems, Pro-ceedings of Adaptive Hypermedia and AdaptiveWeb-Based Systems AH 2000, �2000�, Trento, Italy.

�18� A. KAVCIC, Adaptation in Web-Based EducationalHypermedia With Regard to the Uncertainty of UserKnowledge �Ph.D. Thesis, in Slovenian�, Ljubljana,Slovenia: Faculty of computer and informationscience, University of Ljubljana, 2001.

�19� H. MAURER, N. SCHERBAKOV, Multimedia author-ing for presentation and education: the officialguide to HM-Card, Addison-Wesley, Bonn, 1996.

�20� K. NAKABAYASHI, Architecture of an intelligent tu-toring system on the WWW, Proceedings of the8th World Conference of the AIED Society, �1997�,Kobe, Japan.

�21� K. NG, B. ABRAMSON, Uncertainty Management inExpert Systems, IEEE Expert, 2 �1990�, pp. 29–47.

�22� M. SPECHT, M. KRAVCIK, L. PESIN, R. KLEMKE,Authoring Adaptive Educational Hypermedia inWINDS, Proceedings of ABIS2001, �2001�, Dort-mund, Germany.

�23� J. VASSILEVA, Dynamic course generation on theWWW,Proceedings of AI-ED97- 8thWorldConfer-ence on Artificial Intelligence in Education, �1997�,Kobe, Japan.

�24� G. WEBER, H.C. KUHL, S. WEIBELZAHL, Develop-ing adaptive internet-based courses with the au-thoring system NetCoach, Proceedings of ThirdWorkshop on Adaptive Hypertext and Hypermedia,�2001�, Sonthofen, Germany.

Received: October, 2002Revised: February, 2005

Accepted: May, 2005

Contact address:

Natasa Hoic-BozicDepartment of Information Science

Faculty of PhilosophyUniversity of Rijeka

Omladinska 14HR-51000 Rijeka

Croatiae-mail: natasa�hoic�ri�t�com�hr

Vedran MornarDepartment of Applied Mathematics

Faculty of Electrical Engineering and ComputingUniversity of Zagreb

Unska 3HR-10000 Zagreb

Croatiae-mail: vedran�mornar�fer�hr

NATASA HOIC-BOZIC is an assistant professor at the Department of In-formation Science, Faculty of Philosophy, University of Rijeka. Shereceived Ph.D. degree in computer science from the Faculty of ElectricalEngineering and Computing, University of Zagreb, Croatia. Her mainresearch interests include information and communication technology�especially hypermedia and Internet� in education. She participated inseveral research projects in the field of e-learning.

VEDRAN MORNAR is a professor of computer science at the Faculty ofElectrical Engineering and Computing, University of Zagreb, Croatia,where he currently teaches several graduate and undergraduate comput-ing courses. He graduated and received his PhD degree in computerscience from the same university. His professional interest is in appli-cation of operational research in the real world information systems,database design, development and implementation.

AHyCo: a Web-Based Adaptive Hypermedia Courseware System

Documents

Transcript of AHyCo: a Web-Based Adaptive Hypermedia Courseware System