Nicholas C., Mayfield J. (Eds.). Intelligent Hypertext. Advanced Techniques for the World Wide Web

Springer, 1997. — 196 p. — (Lecture Notes in Computer Science). — ISBN-10 3540636374, ISBN-13 978-3540636373.The papers in this volume grew out of contributions and discussions held at the Workshops on Intelligent Hypertext, which were held in conjunction with the 1993 and 1994 Conferences on Information and Knowledge Management (CIKM'93 and CIKM'94). The theme of the CIKM conference is the interaction between the fields of artificial intelligence, database management, and information retrieval. With their emphasis on intelligent systems for managing information, using the hypertext metaphor, the two Workshops on Intelligent Hypertext were consistent with this theme.
In the years since the workshops took place, the World Wide Web has achieved enormous success. Indeed, for many people, the Web is the only hypertext system they have ever used or even heard of. With this in mind, the editors invited several of the workshop participants to prepare chapterlength papers based on their workshop contributions, with special emphasis on how their results apply to the Web now, or might apply to versions of the Web that appear in the future.
Bieber describes his experience with creating a Web version of the August 1995 issue of Communications of the ACM. He points out that the although rich semantics for hypertext links is a familiar idea in the hypertext community, the Web in its current form does little to support this. In particular, no tool to aid authors in adding semantic information to their links is available, making this process laborious and error-prone. Bieber goes on to describe a browser in which semantic information about a link, e.g., what sort of relationship exists between the current document and the one being linked to, is presented. One of the important lessons to be learned from this effort is that producing a well-designed hypertext document of non-trivial size is still a significant effort.
Brusilovsky discusses adaptive hypermedia, in which the hypermedia system adapts to its individual human users. This idea goes beyond the idea of "preferences," standard in contemporary browsers. Adaptive hypermedia includes adaptive content, in which different users may see different versions of what is essentially the same hypermedia document, whether that document is an ordinary Web page, for example, or perhaps an annotation or index document. Adaptive hypermedia also includes the notion of adaptive navigation, in which some users see links that others do not. Brusilovsky addresses the question of how adaptive hypermedia might be evaluated, and he discusses many example systems.
Gobbetti and Turner describe the Virtual Reality Modeling Language (VRML), and explain how 3D documents may be implemented using VRML and the Web. They present an overview of VRML's historical beginnings, implementation issues, and sample applications. Of particular interest is the "Virtual Sardinia Project," in which geographic information, including satellite images, are used to create a VR model of the island of Sardinia. This model can then be explored with a VRML-enabled Web browser.
One component of intelligent hypertext systems (or indeed of intelligent systems in general) is the degree to which the system adapts (or can be adapted) in order to better serve its users. Kay and Kummerfeld describe their PT system, which builds and maintains a model of each user. For new users, this process begins by PT asking them questions about their background knowledge in computer programming. As the system runs, this user model is updated, and information derived from it is used to generate customized documents, and to customize navigation between documents.
Kent and Neuss argue that World Wide Web search should be carried out by exchange of meta-data, not (as is now practiced) by exchange of the data themselves. They suggest that formal concept analysis can provide tools for representing the semantics of hypertext documents, which can serve as metadata. Conceptual linkage in their formulation is the overlay of relationships among the concepts represented in documents onto an existing hypertext link structure. Conceptual browsing is browsing along the lines of such an overlay. Conceptual browsing represents a midpoint between traversing the links that are part of a document (e.g., by clicking on a WWW link) and using a global indexing scheme (e~g., by searching the WWW with AltaVista or HotBot). The techniques they report apply both to the structuring of local hypertexts, and to guidance of the user through such a structure.
Mayfield presents an overview of two-level models of hypertext. For a hypertext system to be considered intelligent, it can be argued, requires that the raw data be augmented by some meta-data structure. The purpose of such meta-data is to organize the infbrmation so that it can be accessed more easily. Several forms of meta-data are possible, and Mayfield describes a variety of two-level hypertext systems.
In their chapter on the TELLTALE system, Pearce and Miller present their work on the scMability of information retrievM techniques in a hypertext context. TELLTALE is a hypertext system that provides similarity matches between documents (much as traditional information retrieval systems do). But it is integrated with a hypertext interface in a way that greatly improves the speed and ease with which a user can explore the information space. For example, three ~ statistics-based similarity measures are available to the user for controlling how dynamic hypertext links are created. TELLTALE's two main scaling techniques are 1) the use of n-grams (sequences of n contiguous characters) as retrieval terms in lieu of more traditional word-based terms; and 2) the distribution of meta-data about the documents in the collection across several machines. The former adds scalability in the types of documents that may be included, while the latter aids scalability in the number of documents indexed.
The notion of meta-data in hypertext emerges several times in this volume. One specific proposal is presented by Wang, Ghaoui, and Rada, who discuss the use of semantic nets as domain models for hypertext systems. Of particular interest is the issue of how to maintain the consistency (of both meaning and style) of a set of hypertext documents created by many people. As Web content becomes more dynamic, the need for techniques to understand (and control) what material exists, and where it resides, will increase. The MUCH and RICH systems described in this chapter are important steps in this direction.
Rus and Subramanian describe an architecture for intelligent software agents, and argue the suitability of such agents to discover and use information distributed over a network. These agents are intelligent in the sense of being able to examine data, make plans in response to tile data, and perform actions according to these plans. Agents are endowed with "sensors" and are capable of making sophisticated analysis of documents at different levels of detail. The agents are mobile, and migration from one site to another is one possible action. Rus and Subramanian describe the formal underpinnings of their system, and discuss its operation in detail.Enhancing Information Comprehension Through Hypertext.
Efficient Techniques for Adaptive Hypermedia.
Exploring Annotated 3D Environments on the World Wide Web.
User Models for Customized Hypertext.
Conceptual Analysis of Hypertext.
Two-Level Models of Hypertext.
The TELLTALE Dynamic Hypertext Environment: Approaches to Scalability.
Domain Model Based Hypertext f o r Collaborative Authoring.
Information Retrieval , Information Structure , and Information Agents.