[topicmapmail] Application for TM

[Murray Altheim]

Murray Woodman wrote:
> * Carlo Moneti <cmoneti@twcny.rr.com>
> |
> | Hi,
> |
> | A topic map wiki is one application previously discussed as a TM
> | demonstration application. That lead me to think about
> | http://Wikipedia.org. What a public relation coup it would be to turn
> | wikipedia.org into a TM driven site. Wikipedia has done an amazing
> | job. Can TM improve the current system? The software (and data) is
> | freely available: http://wikipedia.sourceforge.net/ . It might be
> | interesting to see how easy/difficult it is to harvest the existing
> | database data into a topic map.
> 
> I have been thinking along these lines as well. The Wikipedia software
> currently supports a "categories" feature which is currently not being
> used. It is intended to be a catch all for whole/part and type/instance
> relationships. I popped by their technical mailing list a few months ago
> and suggested that there would be some advantages if there were a
> mechanism to capture (i) the differences between the above relationship
> types and (ii) the ability to type other relationships.
> 
> It was generally thought that asking users to distinguish between hasA
> and isA was too much of an ask. So, it seems that the main method of
> encoding will continue to be through untyped hyperlinks.
> 
> There is however, a proposal to extend the category system which would
> be very helpful:
> 
> http://meta.wikipedia.org/wiki/Field-value_pairs
> http://meta.wikipedia.org/wiki/Categorization_with_field-value_pairs
> 
> Having said that there are currently "list" pages which would be good
> starting points for starting to build a topic map. I don't know how much
> further you could get using automated means though.

Murray,

I'm not in a position to spend any time with Wikipedia right now,
but if I could make a suggestion, it'd to be to avoid trying a
categorization system for Wikipedia at all, based on fixed categories.
Instead, I'd suggest a Faceted Classification approach, where there
was some way to suggest to Wikipedia editors a place on each page
for them to add a list of either keywords or subject identifiers
(in some form) that the page was "about". Then an engine could
harvest those facets and build a navigation system based on that.
This would have the advantage of being much less brittle in the
same way that FC systems in libraries work: if there are too many
pages that fall under the same categorization (by their accumulation
of facets), they can be further broken down by adding facets to the
pages as differentiating factors, without altering existing searches.

The reason I state this is that the more formal approach (using
either mereological or taxonomical relations) was the road I
originally went down with my own project, but in informal systems
this just doesn't work (I don't think), whereas FC has a number of
major advantages, first of all no need for any real "global" agree-
ment on a classification system. Chris Purcell of the CommunityWiki
is trying out ideas related to this on his FacetWiki.

   http://www.usemod.com/cgi-bin/mb.pl?FacetWiki

Murray (the Other)

......................................................................
Murray Altheim                    http://kmi.open.ac.uk/people/murray/
Knowledge Media Institute
The Open University, Milton Keynes, Bucks, MK7 6AA, UK               .

   An Australian scientist says studying the kangaroo genome might
   help scientists modify genes in cows so that they produce highly
   nutritious milk.
   http://news.bbc.co.uk/1/hi/sci/tech/3604045.stm

   Do we really want this kind of nonsense? What animal has more
   common sense than genetic scientists, such that we can implant
   a gene or two?