Andre Engels wrote:
Why not create [[Mexican]] as a redirect page?
That would solve things for en:, but not for some other languages. For example, in Latin all of the following could be forms to redirect at [[domus]]:
[[domi]], [[domum]], [[domo]], [[domos]], [[domorum]], [[domis]] (and perhaps also [[dome]]). And similar for almost every word. Then a trick like the above would be much welcome, it seems.
Swedish, Danish, and Norwegian are about as complicated as Latin in this respect (Swedish words have 4.5 forms on *average*), while German is a little lighter (perhaps 3 forms per word), and Finnish is a lot heavier (I'd guess 10 forms per word on average).
Still, the free spell checker "ispell" is able to "stem" most forms down to the correct basic word, except for the relatively few ambiguous cases. The existing ispell dictionaries for various languages already contain the information necessary for this.
Using stemming (based on ispell) could be useful both as an automatic URL redirection, and during searches. Commercial "text retrieval" databases do this, but I haven't heard of any open websites or web search engines or free software that use this technology.
For searching, you could either stem each word before you index the text corpus, or you could "unstem" each search expression, so that a search for "domus" actually searches for "domus OR domi OR domum OR domo OR domos OR ...".
The next step would be to "stem" synonyms into "concepts", so that a search for "car" returns hits on "automobile", as well. At some point of generalization, you just get too many hits. So maybe perfect hits should be prioritized over stemming hits, which in turn get prio over synonym hits. Just like title hits get prio over fulltext hits today.
This is a new direction that I have thought about, but never got around to implement. Does anybody have any experience to share?