Enrico Schnepel enrico.schnepel at randomice.net
Do Okt 9 14:32:38 CEST 2008

Hash: SHA1

Hello Torsten,

comments below...

Torsten Ziegler schrieb:
> Dear deepamehta developers
> i've got a short question (feel free to answer also shortly):
> The check field "Unique Topic Names" in the "Topic Type"
> is it used anywhere ?
> I haven't found any implementation for this (yet) in trunk 354
the property describes how the topic map imort handles duplicates. if
it is set to true and an internal and the imported topic with the same
type (I think so) have the same name - they are merged.

> background (longer explanation):
> As some of you might know i am working on
> using deepamehta for explaining the chinese medicine.
> For this I am missing some kind of hierarchical structuring
> of data in DeepaMehta. So right now I am experimenting
> with the implementation of new topic types that allow to
> categorize topics not only by implementing new subtypes
> (as this would have to be done in the Type Builder)
> but by associating them to certain categories (this should be
> done on the "normal" Topic Map)
> And for these features I am also looking in how to make
> topic names unique throughout the corporate memory
> (or at least the part of cm associated to one user, or maybe
> a part of the cm associated to one grand category)
> To illustrate what I am looking for try to imagine this example.
> I think you all know the actors and movies example from the
> DeepaMehta installation.
> If I add to this example a new Information about the role of an
> actor in a movie there will be the three topic types
> -actor
> -role
> -movie
> while it is easy to associate actors with movies it gets complicated
> if you want to specify the role of an actor in a movie.
> Given there are
> actor A -  role 1 -  movie x
> actor B -  role 1 -  movie y
> actor A -  role 1 -  movie z
> actor A -  role 2 -  movie z
> as role 1 (e.g. the gardener) appears several times it has to be
> cloned for each movie, so the example would look like
> actor A -  role 1 in movie x -  movie x
> actor B -  role 1 in movie y -  movie y
> actor A -  role 1 in movie z -  movie z
> actor A -  role 2 in movie z -  movie z
> this leads to a naming scheme that uses very long names
> what I am looking in right now is to
> split the associations to:
> actor A -  role 1
> actor A -  movie x
> role 1 -  movie x
> ....
> and then modify the search for related topics so that
> you can get from
> from actor to role and movie and vice versa,
> but you should be also be able to find all actors playing the gardener
> in any film.
> For sure this needs then a top category attached to it,
> as the gardener has different aspects in movies (e.g. the assassin)
> as in agriculture. But for sure it also has some common aspects,
> so the gardener should be unique throughout the corporate memory.
> If you have any thoughts on this problem
> I am very curious to see other aspects of this problem
> or insights in side effects of my proposal.
This sounds more like the need for a tertiary association. With DM
this is not supported, but it could be emulated with a topic which
connects the actor, the movie and the role.

actor A - play
movie x - play
role  1 - play


Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org


Mehr Informationen über die Mailingliste devel