4.19 Technologies and explicit knowledge continued
The following examples give a taste of what is now making the transition from research laboratories into commercial products. Large hierarchical information structures are extremely common, whether in libraries, organisational charts or websites. Displaying such large structures is a challenge, and since the user soon runs out of screen space, navigating them can be tedious. Screen 7 shows a system that uses animation and carefully designed graphical effects to give the impression of manipulating a three-dimensional structure floating in space. Experiments provide evidence that this kind of interface enables users to browse and search large hierarchies such as library indexing schemes, organisational charts and websites more quickly than using traditional two-dimensional interfaces.
Hierarchies are easy structures to understand. However, other kinds of information have no obvious visual corollary; for example, how does one visualise a library of documents? A large organisation may have hundreds of reports, which typically will be organised only in conventional ways, such as author, date published, project and keywords, and which may be searchable if they are online. However, there are many conceptual links between documents that only become apparent on reading them. Can any of these be automatically detected? Box 4.15 summarises one approach.
Box 4.15 Technology briefing: mapping hidden conceptual structures in digital libraries
Information retrieval techniques are emerging which analyse libraries of digital documents using statistical analyses of terminological frequency and relatedness in order to identify potentially significant clusters of documents. For example, Chen (1999) describes techniques for visualising conceptually related document clusters as maps which can then be delivered over the Web. Such automatic indexing techniques are ideal for large corpora of documents that have no metadata or other form of classification information. Screen 8 is an example generated from a research approach called ‘generalised similarity analysis’. Generalised similarity analysis seeks to identify conceptual structures in a digital library by combining one or more measures: latent semantic indexing (statistical measures of similarity based on keywords in documents), hypertext navigational patterns (reflecting the most heavily navigated routes) and connectedness (for example, by citation or hypertext link). Next, these structures are then visualised using an automatic map layout algorithm, which spatially clusters apparently related documents into branches. Finally, these are delivered over the Web as interactive maps that can be browsed and searched as virtual reality models (using the Virtual Reality Modelling Language).
In the longer term this promises a way to generate automatically visual overviews showing different perspectives of a domain.