[Wikipedia-l] Disambiguation detector

[Lee Daniel Crocker]

> (Toby Bartels <toby+wikipedia at math.ucr.edu>):
> 
> My idea is to link people directly to the disambiguation page,
> which will (if well written) contain exactly what readers need --
> human editing of the disambiguation page in the wiki way will see to that.
> Here's a brief algorithm:
> 
> Given a page title [[X]] without parentheses or of the form [[X (Y)]]:
> * Search for a page [[X (disambiguation)]]
> ** If it exists, add "See also: [[X (disambiguation)]]"
>    (unless "Y" = "disambiguation", of course).
> ** If it doesn't exist, add "See also: [[X]]"
>    (unless we're on [[X]] itself, of course).
> We don't need to link to anything else;
> the disambiguation page (either [[X]] or [[X (disambiguation)]])
> will do so, and furthermore will have human-written text
> explaining the nuances, if necessary.
> 
> As a bonus, we could add the complete list of parenthesised pages
> as a "See also" on the disambiguation page itself
> (as identified by the "unless"es in the previous paragraph).
> That's not really necessary, but it could be convenient;
> for example, it would highlight an incomplete disambiguation page.
> 
> Furthermore, this could solve the problem of formatting disambiguation blocks.
> Larry, mav, and I have discussed this matter before
> on [[en:Wikipedia talk:Disambiguation]] (look for "= Jimmy Carter =").
> If, as mav suggested on the mailing list just now,
> we put this "See also" directly underneath the language links,
> then Larry, mav, and I, at least, should think that it looks good.
> Human-written disambiguation blocks would be unnecessary;
> but the human-written material would still be on [[X (disambiguation)]],
> so Lee (and I) should be happy about that.

That's a bit complex, but maybe it could be made to work. The analogy
to language links doesn't quite work for me because some human editor
has actually gone to the trouble of noting that the fr:chat article is
an appropriate French version of en:cat, whereas these will be fully
automatic. But I agree that it's a reasonably safe bet that X (Y) is
somehow related to X, X (Z), and X (disambiguation), because our
conventions here have created that association, so it's somewhat human-
created.

Another issue is performance. Perhaps we could help both of those
issues by making a single link to a list page rather than a set of
links. But what to call that list is tricky. Is isn't really "related
pages" because they're only related lexigographically, not semantically.
Perhaps it could be something like "Other 'Foo' pages".

-- 
Lee Daniel Crocker <lee at piclab.com> <http://www.piclab.com/lee/>
"All inventions or works of authorship original to me, herein and past,
are placed irrevocably in the public domain, and may be used or modified
for any purpose, without permission, attribution, or notification."--LDC