Knowledge Mapping

[Nick Arnett]

At 01:27 PM 12/18/97 -0800, Linda Hyman wrote:
>Lately, several corporate folks have approached me about "knowledge
>mapping." Does anyone out there do it, or know of anyone doing it
>especially in a corporate environment? When I asked what they meant, here
>is how they responded:
>
>>* Knowledge Mapping -- Basically we're looking for someone with experience
>>     working with database designers to provide the protocols and
>>     attributes for how to define, catalogue,store, search and retrieve
>>     info.

We're experimenting with it.  The idea is that when you can represent
information spatially, you can use cartographic or spatial features to
communication meta-information.  To be a bit more concrete, the kind of
thing we've demonstrated was a layered map-like interface to a Yahoo-style
taxonomy, for displaying search results.  At any layer, you see the
top-level categories and an indication of where there are "hits" in the
taxonomy.  We added pop-up lists for each category and pop-up document
summaries to the lists.  Those features allow you to quickly explore a
results list that would be unwieldy as a typical, flat relevancy-ranked
list.  When you want to drill down, you'd click on a category name and then
you'd see a map of the next level down in the taxonomy.

The advantage is that by displaying a lot of context, people can rapidly
explore and disambiguate.  The latter is vital to systems with lots of
documents when users enter short queries, since they'll usually get a
tremendous number of hits.  The map is a compact presentation of the results
list, with detail quickly accessible.

The downside, or at least unknown, issue is that no one really knows how
people will interpret cartographic features in the context of information
spaces.  What would features such as rivers, mountains, cities, etc., mean
in a map of documents?  Sometimes the "map" can be literal -- search
geographic information and map the results to a map, or search anatomical
information and map the results to anatomical charts.  Those work, but for
the majority of kinds of information, there is tremendous ambiguity.
Imagine, for example, the maps that show colors to indicate rainfall
density.  Is there anything about the choice of colors that is intuitive?
Probably not.

On the other hand, people seem to be very attracted to the spatial metaphor.
Our demo consistently draws a crowd, as some of you saw in New York last
week.  This is not so surprising, considering that "memory palaces" and
related spatial mnemonic devices have been used for thousands of years, only
recently (that is, 400 years or so in Europe, more recently in Asia) falling
out of vogue as a result of the reduced cost of paper.  And although I
suspect the majority of software developers don't know it, icons, windows
and mice were invented to combine motor memory with spatial memory because
they are very effective.  That is, if you move your body and locate
something in a space, you'll remember it much more effectively than if you
don't do either one.

I keep an eye out for interesting research in this area, but nothing much
has really grabbed me.  Perspecta (www.perspecta.com) has one of the better
fly-through spatial interfaces.  For reference to a bunch of others, see
Gerry McKiernan's resource page:

http://www.public.iastate.edu/~CYBERSTACKS/BigPic1.htm

Nick Arnett

--
    Senior Product Manager, Knowledge Applications
        Verity Inc.  (http://www.verity.com/)
        "Connecting People with Information"
  Phone: (408) 542-2164  E-mail: narnett at verity.com