Re: [Evolution] Dictionaries

[Andrew Kephart <akephart agere com>]

Since the request was made for more input, I've been thinking that a 
purely context-based approach might be worthwhile (ok, well, assuming 
that you don't need real-time typo flagging I guess).  It would 
probably be a longer-term feature, though...

The basic idea would be that at each word in the message, the spell 
checker would have an idea of the "expected" language of the word 
(selected from the subset of "active" languages).  For that word, a
dictionary miss indicates either a misspelling or a language change.
The difficult part is choosing which of these two options the checker 
will choose.

        A simple algorithm (which could be used in an as-you-type 
mode) could be:
                (1) found dictionary miss
                (2) search for target in other dictionaries
                (3) if found, set "current" language to language of dictionary
                    where word was found
                (4) if not found in other active dictionaries, flag as
                    an error (keeping the existing "current" language
                (5) move to next word

        A more complex (but probably more accurate, although not really 
workable for as-you-type) algorithm would be to examine the surrounding
words -- the current word would be expected to be in the most commonly 
found language association of some set of N surrounding words (as
determined by dictionary hits of those words).  This would flag the
occasional "ist" for "is", but would play nice in the english-manual-
page-in-the-generally-german-email context.  

        Of course, you could have any number of threshholds and 
algorithmic variations made available through configuration options, no?
(do I need the smiley for that?)

-andrew

> tor, 2003-02-13 kl. 04:48 skrev Ronald Kuetemeier:
>
> > > From my experience that's a problem, specially in none technical
> > settings.  If I write a German business mail, I don't want English mixed
> > in.  Which I sometimes do even after "proof" reading, and if both
> > languages are enabled the spell checker doesn't catch it.  Even if the
> > receiving entity speaks English it can be a bit of a problem. So for
> > business email, none tech, I absolutely prefer _1_ language for the
> > spell checker.  Which should be easy to select from the edit menu,
> > therefore no 1.2.2 for me, except after I port the 1.2.1 behavior
> > forward.   
>
> *Many* Dutch business and technical people have the common, to me
> repugnant, propensity of sprinkling English words and phrases randomly
> into Dutch documents. I've come to the conclusion that they do so to
> look "clever, smart, show one-upmanship." They say: "No, it's just that
> the Dutch are such good businessmen." Yuk.
>
> My Norwegian and English pasts simply make me despise them for not
> taking a greater pride in their own language.
>
> Best,
>
> Tony
>
> -- 
>
> Tony Earnshaw
>
> "Can anyone define 'modern enclitic
> mediocrity' in terms of the Euro for me?"
> - Billy the (Norwegian-Dutch) Cat, Feb '03
>
> e-post:               tonni billy demon nl
> www:          http://www.billy.demon.nl
>
>
>
> _______________________________________________
> evolution maillist  -  evolution lists ximian com
> http://lists.ximian.com/mailman/listinfo/evolution