[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [aspell] Affix or something else
From: |
Björn Jacke |
Subject: |
Re: [aspell] Affix or something else |
Date: |
Thu, 1 Feb 2001 13:09:21 +0100 |
User-agent: |
Mutt/1.3.12i |
On 2001-02-01 at 10:52 +0200 Togan Muftuoglu sent off:
> Now I am not a programmer, can using a morphological approach to the
> words be an alternative to affix compresion ? As far as I understand it
> will work with the root words and make a suggestion based on the
> morphology for an unknown word. Size of the dictionary will be small
> since it will have the root words only.
from my point of view this will not suffice because too many words which
won't exist would be "morphed" and thus many typos would be not
recogniced. The birth of the German ispell dictionary was IIRC when
Geoff Kuenning wrote an affix file as an example for the affix
compression with Ispell. He's no native German speaker and thus he knows
German language much better than a native speaker -- at least from a
technical view, which is important for creating affix rules. You ought
to start from the "root word" and then create affix rules to match the
declined and conjugated forms of the word. This has several advantages
over the automated way of doing this. An automated method of creating
affix rules will never make so high compression results as this manual
method. Furthermore it's much easier to maintain the dictionary because
you just have to add the infinitive of a verb e.g. and know what your
affix flags do to add all forms of this word.
--
Björn Jacke · E-Mail: address@hidden
SuSE GmbH · Schanzäckerstr. 10 · 90443 Nürnberg · Germany
pgpEYl7FLcWNC.pgp
Description: PGP signature