Warning: Parameter 3 to newsxRenderPageBy() expected to be a reference, value given in /data/daniel/brightbyte.de/www/html/mw/includes/parser/Parser.php on line 3333 Ideas for a smarter inter-language link system - BrightByte
Toolbox
  • Printable version
 
Toolbox
LANGUAGES
Language
Personal tools
Categories
Wikipedia Affiliate Button
 

Ideas for a smarter inter-language link system

From BrightByte

Jump to: navigation, search

At the 24C3, I talked to a couple of people about how the inter-language-link system could be improved. Generally, everyone agrees that it's bad that all wikis have to be linked to all other wikis explicitly. For example, if we have an article like Water in 100 languages, 99 inter-language links have to e placed and maintained on 100 pages (so we have O(n2 complexity). That sucks. So, here's a quick sketch of the the idea that formed in my head:

  • Inter-language-links should be maintained at the individual pages, much as they are now.
  • However, the information maintained this way should be centralized in a global inter-language-link table
  • It would be useful if inter-language-links could be inferred from the existing data, like this: if A links to B, and B links to C, A should also link to C. Formally, that means that we want to be able to treat inter-language linkage as a transitive relation, and thus calculate and use the transitive cover.
  • However, it has to be considered that inter-language-links do not necessarily express a direct translation, i.e. the concepts don't have to be equivalent. This is due to the difference in granularity and coverage of subject areas on different wikipedias.
  • A solution would be to consider articles that are interlinked both ways to be equivalents, and to treat this equivalence relation as transitive, and suitable for inferring inter-language connections: if A links to B and B links to A, and B links to C and C links to B, then we can infer that A should link to C and C should link to A. (The question if only explicit links can be used as a basis for inference, or also inferred links can be used for further inference, I leave for later)
  • Thus, we would get two kinds of inter-language-links on each page: the ones that where specified explicitly, and the ones inferred from the global inter-language table. The inferred links would have to be marked somehow, and have an extra link to a page that gives reveals the reasoning used to infer the link: if the connection that was inferred by the system is discovered to be wrong by a user, there must be a way to know how the link was inferred, in order to know how to fix it.
  • In addition, a third kind of inter-language-link could be shown: (explicit) incoming links from other pages.
  • Note that this way, we may get multiple pages linked for the same language - this would require us to not only show the language as the link text, but also the page title. Not sure how to integrate this into the sidebar - maybe only show the titles on hover? But that would probably not be low-tech (read: MSIE) compatible. Alternatively, only the most "strong" kind of link would be shown: explicit preferred, then interred links, and if nothing was inferred, incoming. Though there may still be multiple incoming (or even inferred) links. And even if there is an explicit link established, seeing incoming links may be useful to see when the explicit link should be changed, because a better target became available.

I'm afraid this is a bit dry and technical, please add comments/questions, so I know what to explain better, and how.

Anyway, such a system would provide more information to the reader and editor, while reducing the number of edits required to, ideally, O(n) (i.e. each page only needs to explicitly link to one central page, and that central page would need to link back to them). But the system would be flexible enough to deal with less ideal situations gracefully, too.

There was also talk about having two types of explicit inter-language-links: exact matches vs. similar. There could only be one exact match per language, and it would define the pages as being equivalent, but there might be multiple similar pages linked for each language. That proposal would give editors more control, but it would also allow for inconsistencies between wikipedias. One of the nice properties of the former system is the fact that is allows structure to emerge and adapt, with people managing links in their own wiki, to the best of their knowledge.

Please comment, I'm sure this can still be improved :)

Free Content
Freecontent.png

See also

[talk page]Talk:Ideas for a smarter inter-language link system

[edit] A links to B, and B links to C, A should also link to B

A->B, B->C => A->B is a true statement, though probably not what you meant :)

Indeed. thanks for pointing it out, i fixed it. -- Daniel 01:23, 12 January 2008 (CET)


The above comments may have been left by visitors.

This site's operators can not take responsibility for the content of such comments.