16.1 Motivation

There are different systems trying to assist in information retrieval, orientation and navigation in huge collections of information. One of the best known examples is WAIS (Wide Area Information Server) from Thinking Machines Corporation [Kah91] that has already been discussed in chapter 7.1. WAIS allows full-text search in free-text databases. Although WAIS offers a flexible interface and does an excellent job in quickly searching huge data collections, it does not address some cognitive problems. As a motivation for our approach consider figure I.85 which shows a screen dump of the Macintosh WAIS interface.

Figure I.85 WAIS Macintosh user interface

Automatic generation of structural overview maps:

In figure I.85 users formulated a query in plain English searching for background information on Kenya. Before submitting their query, the users had to decide which databases to search. Based on previous knowledge about the contents of the databases their decision was to search the databases "Atlas" and "TMC Encyclopedia". But how does a first time user know what is contained in which database (arrow 1 in figure I.84). Obviously it would be very helpful to get a self explaining overview or a table of contents of the data contained in a particular database.

Getting an overview of the retrieved data:

In figure I.85 WAIS returned a large number of text files as answer to the query. Unfortunately the answer is presented as a textual list that can easily grow to a few hundred items (arrow 2 in figure I.85). Although WAIS offers sophisticated mechanisms for narrowing or expanding the search, the general problem of representing the retrieved information in an easily understandable way still persists. It would be very helpful to get a graphical representation of the answer that could be perceived at one glance.

The general problem we would like to address in this chapter is the automatic generation of navigational aids like structural overview maps of large collections of unstructured data and the related problem of quickly presenting an overview of retrieved data. In this and the following chapters we present the theoretical foundations of CYBERMAP, the algorithms used for the implementation, as well as sample applications based on actual prototype implementations of CYBERMAP in HyperCard on the Macintosh and in C* on the massively parallel supercomputer Connection Machine CM-2 [Hil87] as well as in C and Java for the World Wide Web.