Information structuring For Intelligent Hypermedia: A Knowledge Engineering Approach

The next generation of hypermedia systems will undoubtedly incorporate many concepts and techniques taken from the field of artificial intelligence and knowledge-based systems, in order to add more intelligence to the basic hypertext model. In this paper, we discuss where this need for more intelligent hypermedia systems comes from, by looking at the major authoring and reading problems characteristic of hypermedia. We then point out the similarity in system architecture which exists between expert systems and intelligent hypermedia systems, and discuss how a knowledge engineering approach to hypermedia information structuring can help us address these authoring and reading problems. Finally, we describe the use of such an approach in our implementation of IKON, a prototype third-order hypermedia system, based on the so-called Model-Map-View-Praxis or MMVP conceptual system architecture. The need for intelligent hypermedia systems The flexibility in the structuring and presentation of diverse types of information that is offered by hypermedia is at the same time hypermedia's greatest asset and its greatest shortcoming. It is hypermedia's greatest asset, since this flexibility allows information to be represented and accessed in very diverse and natural ways. It is hypermedia's greatest shortcoming, since this flexibility makes the information space of a hypermedia system so rich and malleable that its authors have difficulties structuring it, and its readers have problems navigating through it. This imperfection of the basic hypertext model has become apparent in the last couple of years, during which hypermedia grew from a research subject into a commercially viable technology. Moving from small research prototypes to large commercial hypermedia systems has confronted researchers with the limitations of an approach which simply tries to scale-up present solutions to authoring and reading problems. The challenge now facing hypermedia researchers is to shift from hypermedia systems which offer unguided access to static, undifferentiated information to hypermedia systems which provide directed access to changing, specific information. As Halasz [1] has convincingly argued, this requires the development of a next generation of more intelligent hypermedia systems. In this paper, we discuss how we can address this challenge by following a knowledge engineering approach to hypermedia information structuring. Authoring problems: premature organization and system maintenance During the last three years the MIPS group, in cooperation with Elsevier Science Publishers b.v., has built the Active Library on Corrosion (ALC), a CD-ROM based hypermedia system for materials and corrosion engineering [2]. The final system, which went on sale in April 1992, contains well over 3⋅000 nodes of multimedia information and more than 15⋅000 links. As such it is one of the largest commercially available hypermedia systems. During the design and development of the ALC, the two most important authoring problems we experienced (see also [3]) were problems of premature organization and system maintenance. The premature organization problem consists of the fact that as a hypermedia system is being built, the topology of the underlying hypertext network will necessarily have to alter frequently, to reflect changes in the information contents or in the author's ideas about the desirable link relationships in the network. The system maintenance problem consists of the fact that each time changes have been made to the hypertext network, all the nodes and the links between these nodes have to be checked for internal consistency. The present generation of hypermedia systems offers little or no support for this authoring-in-the-large. Reading problems: cognitive overhead and navigational disorientation Amongst hypermedia usability researchers it is now generally accepted that cognitive overhead and navigational disorientation (the "lost in hyperspace" sensation first described by Conklin [4]) are the two major reading problems facing hypermedia users. Cognitive overhead arises when the reader is faced with too much information