Le Mercredi 22 Décembre 2004 17:14, Plamen Gradinarov a écrit :
Baeckeroot alain wrote:
If it is categorized it might be very fast to
extract.
Otherwise it will need to do "very-clever" request, or do some automatic
categorisation ...
I am very much interested in such a solution. Probably we should create
first a special categorial scheme (uniformly prefixed or indexed), link
to the new categories all needed categories and articles, meanwhile
creating new ones, and when everything is verified, apply the
categorisation cut.
Plamen
I not precisely understand your need. For me you want to extract a part of wikimedia
and _you_ know what you want, but it is not in a good shape for beeing searched to.
The problem is to define efficient research criteria, for let say 70% of what
you need eg ( virus & horse & computer ) must give "trojan horse" ;)
I experienced things like that, and it works fine.
More subtle is automatic categorisation, by doing requests and then
adjust the links tables , eventually iterate the process. I ve not
been very far in that direction, but it seems quite promising
(as far as we are clever ;)
Alain