Category : Word Processors
Archive   : HYPELNK.ZIP
Filename : SEARCH

ÚÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ¿
³Search Techniques³
ÀÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÙ

An examination of the many excellent programs on the market reveals that
they tend to have one feature in common:

Most standard programs utilize the ability of the computer to very quickly
hone in on a unique dataset, based on a binary search, pre-sorting, B-tree,
or other indexing methodology, e.g., lookup techniques.

What is missing in the generally available programs is an efficient
browsing technique add-on. This would be geared towards quickly identifying
sets of data that are close to, but not necessarily identical with, a
substantially more complex set of criteria. Note that what we are proposing
differs from the ability to identify topics with, say, several keywords in
the same paragraph. The difference between LOOKUP and BROWSING is discussed
in methods.

Hypertext provides the means to quickly traverse the desired domain,
permitting the user to rapidly find the needed information, but (in a large
text) with no assurance of success or optimum fit. The proposed technique
will make navigation in large hypertexts even easier, assuring that any
relevant text (if it exists) would most probably be immediately found.

What is being proposed is in many ways more ambitious than Traveling
Software's VIEWLINK visualization, or Lotus's MAGELLAN search capabilities.
On the other hand, the application emphasis (at least for now) is on a much
more limited set of files.

The methodology uses analysis and similarity tools,
complemented by preparation programs. These are tied to the specific
hypertext (or other) applications, and to tools for adjusting and
preparing ("prettying") the produced maps for presentation.

The broad similarity concept utilized makes it easier to get around the
synonym problem. Even if a researcher on cognitive maps forgets that a
chick(en) is known as a hen or domestic fowl, the process has a good chance
of finding a desired reference to, say, "Imprinting and Early Learning" (W.
Sluckin, Aldine Publishing, 1965) using other connections in keywords,
topics or cited authors. Furthermore, the maps produced should facilitate
easier (and therefore greater) user involvement. As all of us key- and
glasses-losers know, the best search strategy is one that looks in the most
likely places first and generates a map as it goes along, so as not to
cycle. "And even if one does not find, there is at least the map."

Underlying it all is a mathematical process for selecting a close (as
opposed to exact) data match based on the concept of semantic !.