Enhancing Web support for resource-based learning

by Dave DeRoure, Les Carr, Wendy Hall and Gary Hill

Multimedia Research Group,
Dept of Electronics & Computer Science,
University of Southampton, UK.

Introduction

Successful resource-based learning using hypermedia systems needs a combination of guided tutorials, questions and answers, and simulations, with resource material such as research papers, bibliographic databases and document archives. It is much more than browsing: learners must be both readers and authors in order to carry out constructional activities and to integrate their own work with the corpus of materials. While the Web can support many aspects of resource-based learning, the embedded link approach makes it difficult to reuse a resource, and there is a fundamental distinction between authors and readers.

In this paper we describe ways in which we have experimented with external link databases in the Web, in support of the resource-based development and delivery of learning materials. Our extensions to the Web enable information reuse and republication for different purposes and in different contexts, and they facilitate authoring - i.e. the necessary basis for resource-based learning. This approach is based on our extensive experience with the Microcosm model.

The case for separate link databases

In order to allow arbitrary document types to participate as sources for Web links, it is necessary to allow link information to be stored independently of the data which is the subject of the 'connection'. This is the approach taken by the HyTime international standard for Hypermedia description, and allows links to be made from information in closed data formats (where URLs would not be accommodated), from information on read-only media (such as CDROM data) and from information which the person constructing the link has no inherent write access (e.g. a third-party annotation).

By providing independent links between information resources, different sets of links can be provided for different groups of people for different purposes and at different times. A Web server, receiving a request for a particular teaching document, may check the identity of the client to see if they are a first or second year student in order to decide which set of links to provide with the document (to introductory subject material or to in depth material).

This is a powerful mechanism for information reuse, allowing the republishing, reconstruction and extension of information applications without the usual economic penalties that fixed (embedded) connections imply.

Authoring using Multimedia Assets

The usual style of Web authoring is to write bespoke documentation which combines the constituents of an information presentation with the relationships (links) between those constituents. The documents, often translated from print-oriented format such as wordprocessor files, are monolithic or possibly decomposed into a tree structure. In a single document the author provides a structured and directed argument, defining top, next, previous, up.

However there is often the requirement, especially in developing learning materials, for authoring based not on complex one-off documents, but on 'expensive' individual assets which can be reused in many different contexts. In this environment there is no single meaning of 'up' or 'top', since the 'multimedia document' is a dynamic view on a set of objects.

This kind of bottom-up document construction from resources (as opposed to one-off, top-down document deconstruction provided by tools like latex2html) requires the use of new kinds of link specification, since a single component may have one set of links in one context and another set of links in a different context.

Implementing Link databases

These new link specifications are held in 'link databases' and are used to create Web links between the components of a hypermedia document. These link databases may be interrogated by the server, client or by a separate 'publishing' procedure.

Server-side Link Resolution

In order to present a standard HTML document interface to the client, it is possible to arrange that links are held separately on the Web server, but recombined with the unlinked document into full HTML before they are sent to the client. If a requested document is absent, the server attempts to rebuild it from its constituent parts (the UNIX `make' facility is an example of a very general mechanism for doing this) and the client is unaware of any change in the document.

The extra stage provides some potential benefits for the publishing site:

The links produced by the decomposition program are tied very specifically to the locations (or anchors) where they are found. By relaxing this location constraint such that the link can be located anywhere that the chosen anchor text is found, we can provide flexible links which will move with the text when edited and can be easily replicated to track many mentions of the key word or phrase. By encouraging such flexibility we can see the following added advantages:

Client-side link requests

The previous example showed how a flexible linking strategy could be implemented without recourse to the client application (although how these links are authored from a dumb client is moot). Instead we can postulate that the server documents remain unmodified, and that some responsibility for link following is moved to the client.

This approach requires the same link databases as above, but does not require the server to merge every possible link anchor back into the document before it is sent to the client. Instead, a client helper is used to follow possible links which do not explicitly appear as buttons. In this approach, the reader selects the text which they are interested in and invokes the helper (clicking on a button) to make the request for links.

Authoring Implications

Having obtained the ability to create databases of links which are not fixed to specific documents but which can apply to sets of similar documents it is possible to create "information applications" which are linked together according to the following paradigms
Multimedia asset authoring
a variety of separate information assets can be linked together in various ways
Resource-based authoring
large collections of information and links can be constructed to act as "pluggable modules" to new applications
Task-based authoring
a small number of documents and link databases can be used to prescribe a recommended route through a large suite of documents

Summary

We have discussed the advantages of external link databases for the development and delivery of resource-based learning materials, and described different implementation models. The techniques presented here are currently being used for teaching at the University of Southampton.

References

Dave DeRoure, Les Carr, Wendy Hall, Gary Hill Multimedia Research Group, Dept of Electronics & Computer Science, University of Southampton, Southampton, UK
D.DeRoure@ecs.soton.ac.uk, L.A.Carr@ecs.soton.ac.uk, W.Hall@ecs.soton.ac.uk G.J.Hill@ecs.soton.ac.uk
(c) University of Southampton 1995.