Re: Moving spellcheckers into the Gtk+ stack

[Nathan Fredrickson <nathan silverorange com>]

On Thu, 2006-12-07 at 13:45 +0100, David Nielsen wrote:
> tor, 07 12 2006 kl. 13:31 +0100, skrev Marco Barisione:
> > For long texts it would be good to have an algorithm to recognize the
> > language of a single paragraph, Microsoft Word does it, but I don't know
> > how it works.
> 
> So far as we spell check as each word is entered, if we hit above say
> 90% known in one dictionary then it's probably that language. A good
> guess is easy to make and will often be correct. Naturally this solution
> is far from perfect but it should give us a good indication within the
> first sentence or so. Or am I just spouting bull?

Ordering the active dictionaries by most-recently-hit would minimize
unneeded lookups without the overhead of tracking the language across
sentences and paragraphs.

Nathan