[Wikipedia-l] Re: [Intlwiki-l] LA2

Thomas Hofer th at monochrom.at
Mon Oct 15 12:28:34 UTC 2001

> I joined these two mailing lists (wikipedia-l, intlwiki-l) a week
> ago, and the discussion on translation links inspired me to implement
> this. Whenever I start an article like this:
> 	En katt (engelska: cat) (tyska: Katze) är ett djur.
>   i.e.
> 	A cat (German: Katze) (Swedish: katt) is an animal.
>   or
> 	Eine Katze (Englisch: cat) (Schwedisch: katt) ist ein Tier.
> the words in parenthesis are made into links to that language's
> Wikipedia.

I'm not entirely happy with this mechanism. If there were 15 articles 
about cat in different languages and I wanted to add a 16th, I had to 
insert a link to my new article in 15 different places. I would feel a 
bit uneasy if I had to insert such a link, say, into a chinese page.

In your example, there are three pages which have to contain the 
following information in an *explicit form*:

(1) katt(se)=cat(en) and katt(se)=Katze(de)
(2) cat(en)=Katze(de) and cat(en)=katt(se)
(3) Katze(de)=cat(en) and cat(de)=katt(se)

This is the *real information-content* of the above:

(A) katt(se)=cat(en)
(B) cat(en)=Katze(de)

And this is the way from (A), (B) to (1), (2), (3):

I. Given (A) and (B) the system can *infer* that (C) katt(se)=Katze(de) 
  (following the laws of symmetry and transitivity).
II. And all the other entries are redundant duplicates.

My point is, that the human editors should only have to provide (and 
maintain!) the minimal information (A) and (B) and the *computer* 
should expand it first to the explicit form (A), (B), (C) and then to 
the fully redundant form (1), (2), (3) *at runtime* (i.e. when the 
article is viewed) and *not* by modifying the source-code.

Note to I.) To provide (A) to (C), we only need someone who understands 
(A) swedish and english and another one who understands (B) english and 
german - and that's all. (Note that *any* two of A to C are 
sufficient.) While the explicit form requires that everyone has to deal 
with every language (at some level).

Note to II.) To avoid redundancy is a golden rule of information 
science. Duplicates are a maintainance-hell.

Here's my conception of the implementation: The links to the different 
languages should not be part of the viewable body of the article, but 
part of the headline or footer (which is generated by wiki). A link 
like (A) can be realized using the upcoming wiki-variables (please read 
the posting of Magnus Manske). The variable could be inserted into the 
swedish *or* the english article at any place (tough it would be useful 
to put them all together at the beginning), but they would not be 
directly visible in the text. They only provide the information (A) to 
the system. Then wiki can build the language-alternatives according to 
(I) and (II) and show them in the header or the footer when the page is 


More information about the Wikipedia-l mailing list