[lingu-dev] spell checking problems

[Ken Foskey]

On Sat, 2005-05-14 at 21:51 +0200, eleonora46 at gmx.net wrote:
> Some thoughts to  Ken Foskey's letter:
> 
> > 1. myspell has a poor choice of mispelled words.  Less than other
> > products.
> Myspell has quite a number of suggestions and avoids the 
> ridiculous suggestions of MS Office

To be clearer.  The first word that comes up highlighted appears to not
select the best choice when compared to other checkers.  This is
slightly different to not providing the word as an option.  Note that
this was not measured against another word processor but another spell
checker.

> > 2. myspell appears to use clear text word lists, potentially making them
> > large.  These dictionaries have the same root as other spell checkers
> > but are built differently.
> Clear text word list have a lot of advantages, and IMHO no disadvantages.
> With a secondary flag, as described in http://tkltrans.sourceforge.net/tklspell/secflag.htm,
> they can be made smaller. First the library has to support the 
> secondary flag, and then the dictionaries can be made smaller.

The issue here is that all spell checkers use different implementations
on disk.  This means that there is not enough working together with
other spell checkers.  If at least the root is consistent and munching
the final result is reasonably efficient then we could distribute the
full list and munch it on installation.  The objective of packagers is a
single CD install with maximum impact.

If we do not work on this then a myspell replacement library may be
implemented and then confusion will result with Ubuntu implementing a
totally different spell checker to anyone else.  This might be a good
thing or a bad thing.

> > 3. myspell does not share personal dictionaries with other spell
> > checkers.  We should have a standard place and format that all personal
> > dictionaries are stored.
> For user words it is important to develop logic to be able to use
> the proper flags. For example, lets assume, the word boy/s is in 
> the dictionary. If the user adds toy, it should be added as toy/s in order
> to handle plural correctly.

What I think this means is myspell should pick up all possible personal
dictionaries and all selection and deselection of these personal
dictionaries.

If I add a word into aspell then it should automatically appear in OOo.
If I add a word into OOo then it should automatically appear in ispell
checking.

> > 2. It would be very interesting to ensure that the actual release
> > dictionaries are interoperable as much as possible.  I thought that one
> > of the ispell / aspell clan were going to use the myspell library as a
> > base.  We should actively encourage this.
> 
> Aspell from version 0.6 uses the same dictionaries as myspell. However,
> since no simple editor uses aspell >= 0.6, only emacs and some hard
> installable emailers, the quality of aspell >= 0.6 was not tested yet.

thanks for that..

-- 
Ken Foskey
OpenOffice.org developer