So how to go about it?
  • the content would need to be multilingual: every node would be marked with its language
  • an automatic translator (Babelfish or similar) would present virtual nodes in every supported output language; like the Webster writeups, these would just appear next to any original writeups for the same title
  • an interface would allow users to propose corrections and additions to the title translation table as generated by the translation software (all translation software is horribly inaccurate and incomplete when used on this scale)
  • the Everything software itself emits some text; this would need to be multilingual, too

I'm not gonna do it, but it seems feasible.

Similar support can be used to deal with alternate spelling of words in the same language. See what Everything misses, now and forever.

PS interesting to notice how my writeup is effectively an answer to lakeonfire's, even though it was written earlier, while lakeonfire's is an answer to klash's.