aspell-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [aspell-devel] support for compound words and word skipping by conte


From: Chris Hornbaker
Subject: Re: [aspell-devel] support for compound words and word skipping by context
Date: Sun, 28 Mar 2004 06:53:25 +0000
User-agent: KMail/1.6.51

On Saturday 27 March 2004 02:24, Kevin Atkinson wrote:
> On Sat, 27 Mar 2004, Chris Hornbaker wrote:
> > On Saturday 27 March 2004 01:50, Kevin Atkinson wrote:
> > > On Sat, 27 Mar 2004, Chris Hornbaker wrote:
> > > > Would be great. Of course, I'd have to be able to define what a
> > > > permissible consonant pair is. Also, being able to define /where/
> > > > words can connect at would be a big help in controlling compound
> > > > words.
> > >
> > > By where do you mean controlling which words can be combined?  If so
> > > maybe, if it doesn't get too complicated.
> >
> > I'm not sure. In the word list would allow for the most precision, but
> > I'm not sure how that would affect other things.
>
> I can easily store a tag with the word.  So you can say add rules such as
> a valid compound word is:
>   A + B  but not B + C.
> That can be done fairly easily.  provided there are not too many
> categories.  I am thinking up to 250 or so.

Alright.

(Semi-following the 'special' field's format below)

To start, if I could exclude some letters (spefically ' and y, for this 
language), then that'll pave the way for other conditions.
compound-words
  compound-test-exclude ' y

Then I could define where they generally connect:
  compound-conn CVC/CV *- CVC/C ** CCVCV *- CCVC ** CVC ** CVV ** CCV **

That would be a great start for me and would allow me to say that: The word 
bloti and its short-forms have possible connection points at:
  -bloti
  -blot-
  -lot-
  -blo-
  -lo'i-

and this should exclude articles (like, pavono (140)) from being seen as a 
compound word (technically is it, but none of it fits the compound word 
connection profiles, which allows it to be better handled as run-together 
words, which is most appropriate in that case).

Or I'm thinking out loud and should hush-up. :-)

-- 
Christopher Hornbaker           
Jabber ID: address@hidden     Email: address@hidden
Join the Free State Project!    http://www.freestateproject.org
   "Liberty in Our Lifetime"
Was I helpful? Let me know! http://svcs.affero.net/rm.php?r=Jilks




reply via email to

[Prev in Thread] Current Thread [Next in Thread]