flyspell-textcat

flyspell-textcat uses an external library (libtextcat) to
 automatically categorize the language of the current document.
 Currently this is done on a paragraph basis, which makes this
 minor mode useful for multi-language documents, without parseable
 language tags (such as text files etc.).

 This mode was inspire by flyspell-babel.el by Peter Heslin which
 tries to extract language information from latex commands.

 This mode relies on ispell-multi, which basically caches *spell
 processes for each language seen, instead of restarting new ones
 all the time a languae change is seen.

 [Untested with xemacs]