The Electronic Labyrinth


Most hypertext systems provide a primitive ability to find specific text in the network. This ability becomes important as the amount of unfamiliar text in the hyperbook grows larger. Sophisticated search tools are described below.

Boolean algebra searches use AND, OR, and NOT operators to build a complex search query. Often combined with this is the ability to use wildcard symbols to represent missing letters. For example, "eat* OR ?eat" would match "eating" since the asterisk stands for any number of missing letters. It would also match "peat" (but not "wheat") since the question mark stands for a single missing letter. UNIX users (and programmers of all stripes) are familiar with the GREP utility, which provides for more complex wildcard searches. Other important criteria include those of proximity (eg. "if within ten words").

Morphological search techniques take into account natural language characteristics such as inflection and derivation. Yankelovich describes such a feature available in InterMedia: "[S]earching for the word 'eat' will retrieve documents containing the words 'eat,' 'ate,' 'eating,' 'eats,' 'eater,' and 'eaten'" (138). She compares this method with wildcard searching and concludes that "[u]sers find morphological analysis techniques far easier to understand." However, DeRose has noted that languages other than English are less amenable to searches of this type (190).

Various query languages provide the ability to search using powerful set operations. The Structured Query Language (SQL) was designed for use as an end-user tool for accessing information stored in databases. However, most users find its complexity daunting. Several front ends have been developed to mask this problem; some are subsumed under the moniker Query By Example (QBE).

The ability to name and save search criteria is important if searches are commonly repeated.

The results of a search may be displayed in several ways. The simplest mechanism moves the cursor to the first hit; subsequent occurrences may be found by repeating the search. A more powerful method is to present all of the hits in a KWIC list, from which the user may choose a destination. Such a list might then be used as a filter to reduce the number of nodes the user must view.

© 1993-2000 Christopher Keep, Tim McLaughlin, Robin Parmar.
contact us