Matt Kingston wrote:
Sounds to me like it would be a sort of "expert" system. Consider:
The Wikipedia article Anne Frank could have tags at the bottom like
{WasBorn:1929} (IsA:Woman} {Wrote:The_Diary_Of_A_Young_Girl} {BornIn:Germany}
you get the idea. These facts or relationships tie together different articles in a more structured way. Presumably a user interface system could then be written that would take all the data and be able to do things like list off all the female german authors of the twentieth century.
Since facts aren't tied to one language, the "relationship database" could span all languages (since there is already a system for linking articles on the same subject to different languages). A relationship would have a certain ID internally with "translations" to different languages. So after typing in: {BornIn:Germany} in the english version, the computer would find the french equivelent of the relationship -BornIn- and the french equiv of Germany and update the french Anne_Frank article accordingly.
Anyway, I'm no specialist in the area, I do know that there are some internet projects of this sort trying to teach computers "common sense" {Lion IsA Animal} I don't know how much success they've had. It might be usefull, but I'm not so sure it fits into wikipedias mandate.
Just my three cents.
Matt
I'm working on something like this already; it's an experimental Wiki with the ability to embed special links, in this case with the format [[property = value]]. These are then used to maintain entries in a tuple table, which is in turn used to generate an alternate RDF view of the page.. Yes, you can do [[is a = thing]]. Bits of Web GUI are generated based on class deifinitions, and you can also do things like import CSV tables, and I hope export them as well. I'm currently working on [[same as = thing ]] and logical inference, as well as extending this to trans-wiki inference.
Hopefully some of this experience will be useful, when it becomes time to add this to Wikipedia. For now, it won't hurt at all to use templates to encode this sort of information, so some data will be available first, before the knowledge-mining software gets written.
-- Neil