9.6 SuperBook - Automatically Structured Documents

SuperBook has been developed for the last few years at Bell Labs [Rem87]. SuperBook preprocesses on-line text and automatically generates a dynamic table of contents, text pages and a history of search words. Contrary to most hypertext systems, which are best suited for authoring new information, SuperBook is well qualified to postprocess existing documents by adding additional navigation and information exploration mechanisms. The original goal of SuperBook was simply to improve the usability of on-line text.

SuperBook takes as input formatted text, e.g., in troff, Scribe, or Interleaf format, and produces three navigation tools:

• A dynamic table of contents is generated based on the formatting macros of the text processing system. This does not mean that a new document structure is being computed automatically based on the document's contents, but rather that the original hierarchical document structure is being replicated electronically. This electronic table of contents allows a user to jump to any subchapter by mouse click.
• A full-text index is built automatically allowing the user to search for any word, word stem or Boolean combination thereof. The table of contents is synchronized with the keyword search in the sense that it automatically displays the proportional number of matches in each subchapter and thus provides an additional method for the user to locate quickly areas of interest.
• Because the ideas and concepts the user is searching for are rarely expressed in the same words by user and book author, an is provided, allowing the user to add index synonyms. The idea is, of course, to collect a "repertoire of synonyms" built up by a community of users.

SuperBook has been implemented in a two-level architecture, isolating the text processing package-specific part into the lower layer. By providing an additional module for a new text processor, SuperBook can thus be adapted to process documents formatted in any text processing package. In a comparative experiment for the Chemistry Online Retrieval Experiment (CORE) [Ega91] researchers found the SuperBook implementation superior not only to the paper version, but also to more conventional electronic document retrieval systems.

Figure I.40 SuperBook Proceedings of the 1993 International Workshop on Applications of Neural Networks (see part IV for a brief description of these proceedings) http://superbook.bellcore.com/SB/IWANNT/iwannt93.eprocs.html

SuperBook is not a true hypertext system because it does not offer explicit links between hardwired link anchors. Rather it provides excellent navigation facilities and allows "linking by query formulation", a process similar to dynamic linking.