ADILE: Architecture of a Database-Supported Learning Environment.

This article proposes an architecture for distributed learning environments that use databases to store learning material. As the layout of learning material can inhibit reuse, the architecture implements the notion of "separation of layout and structure" using XML technology. Also, the architecture proposes to separate the presentation layer from the actual computer application that retrieves ULMs, creates menus and forms, and facilitates navigation. Only at the last moment, is layout added. This allows a learning environment to present itself in many different layouts, enhancing possibilities for reuse of learning material. Also, the presentation layer can be implemented using different technologies; the same application can both present itself through the World Wide Web (WWW or Web), as well as on the desktop of a computer for off-line learning. A prototype has been built, and proves that our ideas are implementable and yield a learning environment with a very flexible layout.

Web-based or web-supported teaching and learning is a fast developing application area of information and communication technologies. The Web was originally designed for presenting static hypertext documents, but during the past few years the Web is used more and more as a vehicle for interaction with a myriad of applications that have more functionality than just to present hypertext documents. These applications can be written in script languages such as Common Gateway Interface (CGI), Pen or Hypertext Preprocessor (PHP), as Java servlets or as webserver modules. These software components provide dynamic content instead of static documents. The dynamic content can contain, for example, the ever-changing contents of a shopping cart, learning material specifically composed for a particular learner, or an authoring environment in which teachers can prepare course materials and publish them for their students.

The applications that provide the dynamic content often make it also easy to compose a website, to create (templates for) pages, to design interactivity, and to make sure that the pages link properly to each other. Often, a database is used to store the pages or the information from which the pages are generated (such as shop inventory listings). Software components between the web server and the database create the actual HTML pages and the links between them. This is already common practice in the business world. For educational purposes, however, there are some problems that need to be solved first. What is the nature of the learning material that is stored in the database? How can this information be retrieved again, and formatted into deliverable materials? How to make it as easy as possible to reuse this information in different educational contexts? How to deal with the special kinds of interaction between learner and learning environment?

In this article, the author tries to provide answers to these questions. In particular, the focus is on aspects of reuse in different contexts: what makes these contexts so different that learning materials cannot or will not be reused? How can we approach these differences be approached to reduce their impact on the reusability of learning materials? To answer these questions, the concept of an educational database needs to be considered.

EDUCATIONAL DATABASES

In the 1970s and 1980s, learning environments were often built using authoring software (Clark, 1985). In these software packages, the user could use multimedia objects, such as audio clips, movies, and pictures by importing files that resided on the local hard disk. This method of creating multimedia learning material can be quite expensive (Tan & Nguyen, 1993), so that some researchers focused on reusing multimedia learning material by storing learning material in a multimedia database (Persico, Sarti, & Viarengo, 1992) and providing search facilities so that learning material could be easily retrieved. The material is entered into the database in the form of multimedia objects, and often metadata is added to it for retrieval purposes. Metadata means data about data, such as: who created it, for what purpose, what are the learning objectives, and what pedagogy is used. These characteristics have to be entered by a human being, as a computer is not able to automatically fill in this data.

How did teachers share learning materials before the age of networks and databases? The multimedia files resided on the hard disk of the teacher who created them. Exchanging these files by floppy disk was often not possible because of large file sizes. A simple improvement would be to connect the computers together by way of a local area network (LAN), so teachers could share each other's files easily. However, searching other people's hard disks for certain materials can be difficult. Every person uses their own organizational structure, and files are only identified by their filenames; this can be very cryptic.

A better way to store these multimedia objects is to put them into a database in a structured manner. Educational metadata tags can be added to them, so that a teacher who is looking for something can better judge the contents of the multimedia object without having to study the entire contents. The multimedia object (that is created for educational purposes) plus the educational metadata will be called a learning object.

A database allows a user to specify a query (written in a query language such as Structured Query Language, SQL) or to specify search criteria on a computer form. The database system can rank the search results in some kind of order, allowing the user to select the most relevant ones. The database system would then provide the following advantages (see also Elmasri & Navathe, 1989; Adjeroh & Nwosu, 1997):

* access control on the objects: only "owners" of a learning object can change or erase it; others can only read or view it;

* consistency between learning objects, so that references from one object to another are always valid (e.g., a database system could enforce the constraint that an object cannot be deleted if a reference to it exists); URLs that point to an object (within the database) are then always valid;

* multiple views on the same data, allowing a teacher, for example, to see all objects, while a learner can only see the objects that fit her or his learning profile; and

* separation between program and data, so that a programmer works away from concrete data or specific cases. If data (learning material) is interwoven with the program, then the help of a programmer is needed to change the data if the teacher wants to change some details of the learning material. If the program and the data are separated, then the data can be changed safely without breaking the program.

The widespread use of commercial Database Management Systems (DBMSs) indicates the value of DBMSs in organizing the storage of data. The educational field is not an exception.

REUSABILITY

In earlier research, the need for reusable components for educational software was identified (Rada, 1995; Sarti & Van Marcke, 1995; Olimpo, Chioccariello, Tavella, & Trentin, 1990; Verhagen & Bestebreurtje, 1995; Bestebreurtje & Verhagen, 1992). The primary motive for focusing on reusable components (such as learning objects) is to reduce the development costs of educational (especially multimedia) learning material. However, there are several factors limiting the possibilities for reuse of learning material.

Educational factors. Reuse can be useful when two teachers, or two courses, share some common subject matter. If there are no commonalities, then there may be very little opportunities for reuse.

Economical factors. Creating online (multimedia) learning material can be quite expensive; so the institution that paid for it, is highly motivated to reuse it within the institution, but the institution is not eager to allow other institutions to reuse it for no cost. Currently, e-commerce technology is not sufficiently integrated into everyday online life to permit institutions to easily pay for reused materials; this makes it more difficult to reuse non-free materials.

Technical factors. Learning material that is to be reused has to be found first. Finding precisely what one needs without being overwhelmed by a great deal of irrelevant material is often very hard. This is even harder when multimedia objects are involved: the current Internet search engines (and their algorithms) primarily index keywords that are present in the (textual) contents of an object; this is not possible with audio and video objects. Often, even documents that are not written in HTML or plaintext, such as Microsoft Word, PowerPoint or PDF documents, are not indexed.

The area of information retrieval (IR) in computer science is involved in researching ways to retrieve precisely what one is looking for (Wold, Blum, Keislar, & Wheaton, 1996; Faloutsos Equitz, Flickner, Niblack, Petkovic & Barber, 1994; Brown, Foote, Jones, Young, & Jones, 1995). However, most of these techniques are not sufficiently mature yet to be useful.

Sociological factors. A well-known effect that inhibits reuse is the "not made here" syndrome (Olimpo et al., 1990). This means that teachers are less eager to reuse learning materials made by "the competition."

Another factor that can play a large role, is the "voluntary labeling problem": learning material that is stored in a database is often manually labeled with metadata. Entering it is often a voluntary task that means extra work. So, often teachers are trying to skip the labeling task, or are doing it very superficially. As the metadata plays a crucial role in effectively retrieving learning material (Hiddink, Peet van der, Verhagen, & Blanken, 2000), erroneous metadata (e.g., entered by unmotivated people) will have a negative impact on the reusability of the learning material.

Legal factors. The Web offers a very large collection of potentially reusable online courses; however, copyright constraints limits the reuse of the materials of these courses. It is not acceptable to copy online materials from "foreign" sites without permission; one can link to these materials from one's own course's websites. This introduces the risk of links becoming invalid, or the page contents being changed unknowingly.

In the previous text we have spoken about "learning material"; but how can we store learning material in a digital format in a database? This question will be discussed in the next section.

LEARNING OBJECTS

Since the early nineties, researchers have begun to explore the concept of learning objects that are stored in a database, recognizing the need to reuse some kind of educational components (Chalon, 1994; Olimpo et al., 1990, Persico, Sarti, & Viarengo, 1992). For example, Persico et al. (1992, p. 214) defined the concept of a Unit of Learning Material (ULM) as follows:

a ULM is an abstraction which collects a variety of segments of learning material based on different media and featuring different layouts, languages and encoding types, with their differences concealed to offer a common and consistent interface to its clients. A ULM may be a simple monomedia object or a composition in space and time (layout) of several pieces of learning material; its multimedia message to the user possibly includes texts, still frames, dynamic sequences and digitized audio.

Compared to most earlier models, this model of a ULM differs on a number of characteristics, which will now be discussed.

* Sometimes, researchers propose to use size limitations on ULMs. However, size restrictions may annoy the teachers. If a teacher wants to enter some learning material into a database which s/he thinks really is a unit that should not be broken into pieces from an educational viewpoint, then the system should allow such a component to be entered into the database at all times.

* A teacher should be able to take a collection of ULMs, and recombine these into one large unifying ULM. This new ULM should be no different from other ULMs; in other words, the model of ULM is recursive: a ULM can contain ULMs, which can in turn contain ULMs, and so forth. This allows a teacher to reuse learning material at any desired level of granularity.

* The ULM not only contains learning material, but also its history (who has used it in what course for what target group), and educational characteristics such as the target group (classified in age or classes), school type, pedagogy used in the ULM, an so forth. A teacher should be able to search ULMs based on these characteristics, formally specified as metadata. The concept of educational metadata is becoming more and more important for online educational databases. Currently, an educational standard for metadata for learning objects is in preparation by the Learning Objects Metadata Group (1484.12) of the IEEE (2001).

* It is considered vital that the layout be not stored in the ULM, and that the ULM is as independent of the layout as possible. If the ULM depends on the layout, then it cannot "live" without it, and it will not be possible to reuse the ULM with another layout at other educational institutions or faculties that use a different style on their web pages.

* A ULM is connected to other ULMs through conceptual links, so that users may navigate a conceptual network with learning material at the nodes. Thus, learners may navigate through what can be seen as a "knowledge landscape" (Bestebreurtje & Verhagen, 1992) or a concept map (Reigeluth, Merrill, & Bunderson, 1978; Hendley, Whittington, & Jurasheck, 1993; Sarti & Van Marcke, 1995).

The model of a ULM is shown in Figure 1. The figure shows a content part, in which movies, audio, and text can be present. In the other part of the model, the metadata, relations to other ULMs, and the history record is shown.

Although we have stated that the size of the ULM is unrestricted, it should be noted that small ULMs are often more generic, and therefore probably more reusable. Large ULMs may contain more details, which makes them less reusable. Yet, extremely small ULMs (e.g., individual sentences) are less usable, because one would need many of them to compose learning material of a reasonable size. More effort is needed to arrange these small ULMs, and to make sure that they connect well to each other. The best size for ULMs is a trade-off between reusability and usability.

The primary goal of adding metadata is to enable effective searching through the database. The metadata enables the teacher to search for very specific characteristics (i.e., amount of interactivity, target group, educational level), or to define a measure of relevance to the teacher (Hiddink et al., 2000).

SEPARATING LAYOUT FROM CONTENT

Many teachers enhance their courses using web sites which are very accessible to other teachers who may want to reuse the materials they find on the Web. However, the markup language that is used to build these pages, HTML, has become very layout-oriented since its inception. Many web pages use HTML to specify fonts types, font sizes, colors, placement of figures, use tables just for aligning text or to create color bars, and so forth. This presents a major limitation to reuse: a teacher may like the contents of a particular web page, but dislike the layout of it. Often, an educational institution has its own layout style for web pages, so "foreign" material may not properly fit in.

It is proposed to solve this problem by separating layout from content, by using the eXtensible Markup Language (XML, see Microsoft, 1994) to markup the structure of the content. The layout is added only afterwards using the eXtensible Stylesheet Language (XSL). Using XML, the structure of the learning material can be described with it's own markup-tags.

Crash Course in XML

In XML, blocks of text are tagged at the beginning with a [less than]tag[greater than] keyword, and at the end with a [less than]tag[greater than] keyword. A tag can have attributes that tell important information about the tag. For example, the element ULM in line 1 in the next code example has an attribute id that tells us that this is ULM number 12553. A tag that is just there for its attributes but doesn't really markup text, has a "begin and immediately end" format: [less than]tag/[greater than]; see the video object in line 16.

This may look as a frightening piece of computer code, but upon closer study it is not so difficult. The code tells us that, globally, a ULM consists of content (lines 2 to 18), relations (lines 20 to 24), history (lines 26 to 28), and metadata (lines 30 to 35). The content, in our case, consists of an introduction (lines 3 to 9), theory (lines 10 to 14), and an example (lines 15 to 17). The example consists of an mpeg movie (line 16). There is a relation to another ULM (with id 12554) of which this ULM is theory (line 21), so the other ULM can be regarded as an example for this one. The name of the relation link is indeed "example." The course has been used in the Telelearning course, Fall 1999 (line 27).

Now let's examine how layout is added to this structure.

Adding Layout

Layout is added to XML using the eXtensible Stylesheet Language(XSL). XSL allows one to specify that the title of the tag "Introduction" should become the title of a new HTML page, and that the body of the HTML page should contain the text of the introduction. We can specify that the contents of the head tag within the introduction should be printed in anal font face, color red, big and centered. The text should be printed in normal face, color black. Just as an illustration, the XSL code to print the head of the Introduction is given:

So, using XSL we can add layout to the ULM without affecting its actual contents. This goes far beyond changing colors and font faces; the entire layout of a website can be encoded in XSL by generating HTML frames, tables, navigation bars, and so forth. This means that the ULMs can be reused in contexts with different layout demands without change, thus increasing their reusability.

The Teacher and XML

But does this mean that a teacher has to write XML code? No, but learning environments can work with XML behind the scenes to create ULMs that are independent of their layout, and that can be reused easily. It is even very easy to change the layout of ULMs if the educational institution decides to change its house style using new features of the next version of the HTML standard and new navigation techniques, or because the institution has merged with another one, adopting a new house style. Using XML, the look-and-feel of a website can be altered without rewriting the contents of the website.

To enter XML-formatted materials, the teacher can use commercially available XML editors. The materials are then uploaded into the database. This works similar to authoring HTML pages using authoring tools, but there is one drawback: as content and layout are separated, the teacher may not be able to see directly how the page will look like to the student, which is important to some teachers.

ARCHITECTURE OF AN EDUCATIONAL DATABASE

What does a system look like that is able to handle ULMs that are actually XML documents, and that can create HTML pages on the fly? Before answering this question, some other requirements that such a system has to meet will be summarized.

Requirements

The educational database system should allow the user (instructional designer, teacher or student) to perform the following tasks:

* enter learning material in the form of ULMs, which may be written down as XML documents. Entering data may be implemented as uploading a file, referring to a URL for inclusion, and so forth. An educational database should be separated from authoring activities.

* add metadata to ULMs that have been entered. The database system should also allow the user to change metadata of her/his ULMs.

* search the database for ULMs, using the metadata to see if ULMs match the user's requirements. There should also be a simple keyword search interface.

* create a course schedule to configure learning routes through the collection of ULMs for a certain course (Verhagen & Bestebreurtje, 1994). The system should also allow the teacher to create a 'topic zone' to which students with a particular search assignment have access. Using such a zone, a teacher can precisely define in what region of the database the students can search, so that they are not confused by irrelevant topics.

* add layout to ULMs using the HTML language, so that students can view the learning material using their web browser. The system should also provide the educational institution's default style, so that teacher does not have to write HTML code.

* provide navigation so that students know their position in the learning route and where they can go next.

Classical Architectures

Classical architectures to make database contents accessible by way of the Web use so-called server-side scripting, which means that special HTML pages are written with embedded database statements. The web server is extended with functionality to interpret these statements and executes them each time a user requests the page. The communication with the DBMS is mostly done using Open Database Connectivity (ODC), which is a programming interface for communicating with databases.

The results of the database statements are then inserted into the HTML page. Examples of script languages are Hypertext Preprocessor (PHP) and Active Server Pages (ASP). The disadvantage of this architecture is that the database queries are intertwined with the HTML code, which makes it difficult to maintain each of them; parts of the program for retrieving learning materials (the database statements) are interleaved with the presentation code (the HTML tags). If a new architecture for presenting learning material by way of the Web is to be developed, then solving these problems might as well be done also.

ADILE Architecture

An architecture for an educational database system has been developed that is able to fulfill these requirements (Figure 3). It is called Architecture of Distributed Learning Environments, ADILE (2001).

The system consists of a database and an accompanying DBMS. Examples of commercially available DBMSs are Oracle or Microsoft SQL Server; but there are also freely available DBMSs, for example MySQL.

The core component of the system is the ULM manager, which interacts with the DBMS to retrieve ULMs on the bottom side, and which interacts with a presentation layer on the top side to send the ULMs to the user. The ULM manager contains all programming instructions necessary for communicating with the database, for creating online menus, forms, and so forth; but also for retrieving course contents and sequencing ULMs. It is a complete computer application in itself, with the difference being that it does not use a computer screen as output, but it uses XML (and HTML) as a transport medium to communicate with users anywhere on the Internet. So the entire system should not be seen as a collection of web pages, but as a computer application with a remote user interface. This paradigm is not new; the X11 windowing system for the Unix Operating System (and derivatives) has been capable of doing this for more than 15 years.

The ULM manager communicates with a Presentation Layer that is responsible for getting the interaction across to the user. If the Web is to be used to communicate with the user, then the Presentation Layer consists of a Presentation Server (that can be a web server) and a Presentation Client (that can be a web client). The Presentation Server translates all XML code into HTML pages by adding layout (in the form of XSL stylesheets), and transmits it across the Web. The Presentation Client receives the HTML pages and displays it to the user. There is also communication the other way around: online forms that the user has filled in are sent from the Presentation Client to the Presentation Server, which translates them back into XML code.

One might ask, why use XML instead of HTML? The answer is that it is thought that XML is more powerful than HTML; it is extensible, so more interaction messages can be added if needed. Also, it is desirable to be independent of the Web: if the learning material involves a lot of video, then the Web may be too slow. A Presentation Layer that doesn't use a network at all may have to be created, but instead a Presentation Client that is nothing more than an application that inns on the same computer as the ULM Manager could be developed. In that case, no data needs to be sent across a network at all, so that video learning material may be used. The database can, for example, be put on a CD-Rom that is given to the students to use at home.

The Presentation Server and Client can also be integrated into a Java program that translates the XML code into Java windows calls (using the Abstract Window Toolkit), so a student can also run it independent from the Web. The Java Presentation layer could provide a more fancy user interface than a web browser; or it could implement multimedia players so that the student is freed from the hassle of installing a list of plug-ins before being able to see the course contents.

Let's look at the ULM Manager in some more detail. Figure 4 shows what components the ULM Manager is built of: the Interaction Processor processes interactions with the user (which menu option was selected, what values were filled in which form, etc.). The processor knows all menus, all forms, and knows what method (Java program code) to invoke when a certain menu option or form was submitted. If necessary, the Interaction Processor invokes methods from other components: the Course Player, the Course Editor, or the Encyclopedia. The Interaction Processor also adds the output of the Navigation Manager and the History Manager to the screen. The Navigation Manager provides global navigation through the program, and the History Processor provides an overview of the steps the user has taken in the past; the user can go back to any one of these steps.

The Database Abstraction Layer provides a uniform programming interface that abstracts from the database manager to be used; it contains all database-specific details such as the precise syntax of the database queries, how to connect to the database, how to receive query results, and so forth. If the developer chooses a different database management system with a different programming interface (e.g., not compatible with ODBC), then only this layer needs to be rewritten. The advantage of this architecture is that one application, the ULM Manager, can handle all these different situations without having to be rewritten.

Also, each component in the architecture has its own specific function, and does not have to bother about other concerns. For example, the ULM Manager does not know how the actual menus and forms are displayed, which makes it easier to write the program; the programmer can just focus on creating the application itself.

The teacher and the student also benefit from this architecture: the teacher can use the same ULM in his/her online courses, as well as on CD-Roms that are handed to the students to study at home. The teacher can also re-use ULMs that were created at a different educational institution with a different house-style, and add her or his own layout to the ULMs. The student will then experience the learning material as more consistent; navigation will also be identical within all learning materials.

A disadvantage is that the web server has more processing to do: instead of just retrieving a file from the hard disk, it has to process an XML to generate an HTML document. Due to this extra load, longer response times can be expected.

Prototype

The author wrote a prototype that implements the ADILE architecture previously described [less than]http://www.wwcn.org/-grit/dil[greater than]. It used only freely available software components: an Apache webserver (as Presentation Server), a MySQL Database Management System, the Cocoon XML publishing framework to generate HTML pages from XML documents, and an Apache Java Servlet Engine. The ULM Manager is implemented as a Java servlet; a servlet can be seen as an auxiliary server that assists the web server in handling certain types of requests. The prototype allows the user to switch to another website layout (including navigation, button style, the use of frames) with the click of a button.

DISCUSSION

The architecture has proven to be very useful in designing and developing educational database systems. It is indeed very simple to create a new layout for the entire learning system in a couple of hours, without even changing the programming instructions.

Using a stylesheet for things like menus, forms, and navigation also makes sure that the appearance of the website is consistent, that is, all pages have the same look and feel to the user, both student and teacher. This has another advantage: style changes are made at one central place, which affects the entire site automatically; no pages can be forgotten.

Recently XHTML 1.0, which is a "marriage" between XML and HTML, has become an official W3C [1] recommendation, and as such can be expected to be supported by the next generation of web browsers. Also, web browser software companies are already beginning to include XML support into their web browsers. The author therefore thinks that the use of XML in the learning environment fits very well into the future of the Web.

There are motives, however, to not remove the layout from the structure of ULMs: some educational institutions consider layout styles their "watermark," so that everybody can see who originally created the learning material. This effect assures that the quality of the learning material adds to the institution's reputation. Separating this watermark from the content disables this effect.

Another issue discussed is that separating layout from content is not WYSIWYG [2]: you will not always be able to see what you will get. Writing appropriate XML authoring tools can solve this problem.

Notes

(1.) World Wide Web Consortium, the organization that develops web protocols to promote interoperability [less than]http://www.w3c.org[greater than].

(2.) What You See Is What You Get, which means that a computer program displays text and graphics exactly as it would look like when printed or published on the Web.

References

University of Twente (January 10, 2001). Architecture of DIstributed Learning Environments (ADILE) (v0.5). [Online]. Available: http://www.wwcn.org/-grit/dile

Adjeroh, D.A., & Nwosu, K.C. (1997). Multimedia database management: Requirements and issues. IEEE Multimedia, (3), 24-33.

Bestebreurtje, R., & Verhagen, P.W. (1992). ODB Project: Developing the concept of an instructional multimedia database for multiple target groups. Presented at AECT '92 conference, Capture the vision, February 4-9, Washington, DC.

Brown, M.G., Foote, J.T., Jones, G.J.F., Young, S.J., & Jones, K.S. (1995). Automatic content based retrieval of broadcast news. In Proceedings of the third ACM international Multimedia conference, pp. 35-43. San Francisco, CA.

Chalon, R. (1994). The DISCOURSE production model of interactive multimedia. Presented at the Multimedia and distance learning for science and technology conference, Rome, 28-29 March. [Online]. Available: http://www.irpeacs.fr/papers/rc/discrome.htm

Clark, R.E. (1985). Confounding in educational computing research. Journal of Educational Computing Research, 1(2), 21-29.

Elmasri, R., & Navathe, S.B. (1989). Fundamentals of database systems. Redwood City, CA: The Benjamin/Cummings Publishing.

Faloutsos, C., Equitz, W., Flickner, M., Niblack, W., Petkovic, D., & Barber, R. (1994). Efficient and effective querying by image content. Journal of Intelligent Information Systems, 3(3/4), 231-262.

Hendley, R.J., Whittington, C.D., & Jurascheck, N. (1993). Hypermedia generation from domain representations. Computers and Education, 20(1), 127-132.

Hiddink, G., Peet, G. van der, Verhagen, P.W., & Blanken, H.M. (2000). Increasing retrievability and reusability of learning material by developing a measure of relevance based on academic teachers' conceptions. Presented at the AECT 2000 International Convention, Long Beach, CA. [Online]. Available: http://www.wwcn.org/[sim]grit/publications/aect92.html

IEEE Learning Technology Standards Committee (LTSC) (January 10, 2001). Learning Object Metadata Working Group [Online]. Available: http://ltsc.ieee.org/wgl2/

Microsoft (1994). Microsoft open database connectivity software development kit version 2.0. Redmond, WA: Microsoft Press.

Olimpo, G., Chioccariello, A., Tavella, M., & Trentin, G. (1990). On the concept of reusability in educational design. In Learning technology in the European communities. Proceedings of the DELTA conference on research and development, pp. 535-548. Dordrecht: Kluwer Academic Publications.

Persico, D., Sarti, L., & Viarengo, V. (1992). Browsing a database of multimedia learning material. Interactive Learning International, 8, 213 - 235.

Rada, R. (1995). Developing educational hypermedia: Coordination and reuse. Norwood, NJ: Ablex Publishing.

Reigeluth, C.M., Merrill, M.D., & Bunderson, C.V. (1978). The structure of subject matter content and its instructional design implications. Instructional Science, 7(2).

Sarti, L., & Van Marcke, K. (1995). Reuse in intelligent courseware authoring. In: N. Major, T. Murray, & C. Bloom (Eds.) AI-ED-95 Workshop on authoring shells for intelligent tutoring systems, Washington, DC, 16-19 August. [Online]. Available: http://www.pitt.edu/[sim]al/aied/sarti_vm.html

Tan, W., & Nguyen, A. (1993). Lifecycle costing models for interactive multimedia systems. In C. Latchem, J. Williamson, & L. Henderson-Lancett (Eds.). Interactive multimedia: Practice and promise, pp 151 - 164. London: Kogan Page.

Verhagen, P.W., & Bestebreurtje, R. (1995). Towards the architecture of an instructional multimedia database. Journal of Computer Assisted Learning, 11(3), 80-91.

Wold, B., Blum, T., Keislar, D., & Wheaton, J. (1996). Content-based classification, search, and retrieval of audio. IEEE Multimedia (3), 27-36.

Acknowledgements

The author wishes to thank P. Verhagen for his valuable comments during the first version of this article and Jan Stefanides for doing the SGML to XML translation.