Re: [Apertium-stuff] free Zulu analyser - hunspell also for English

[Francis T. <ft...@pr...>]

El dj 04 de 06 del 2009 a les 21:31 +0200, en/na ge va escriure:
> >We have our own transducer for analysis/generation
> >of English.
> 
> I see. So apertium has a complete own transducer, that
> has nothing to do with jtoolkit or other free tools.

Exactly.

> You call transducer a pos tagger, lemma identifier and generator
> in this case.

In this case a transducer is a morphological analyser / generator. The
part-of-speech tagger is a separate program. 

> generator:
> if I say be <past><pl><p3> it says: was
> if I say be <present><sg><p1> it says: am
> 
> Can I invoke just the generator somehow?

$ echo "am" | lt-proc en-ca.automorf.bin 
^am/be<vbser><pri><p1><sg>$

$ echo "am" | lt-proc en-ca.automorf.bin | apertium-tagger -g
en-ca.prob 
^be<vbser><pri><p1><sg>$

$ echo "am" | lt-proc en-ca.automorf.bin | apertium-tagger -g en-ca.prob
| lt-proc -g ca-en.autogen.bin 
am

For further details see http://wiki.apertium.org/wiki/lttoolbox

> Hunspell is also fast, when you use chmorph/analyzer.
> Only read in of the dictionary (hunspell class creation) is slow.

But lttoolbox is faster, much faster.

Fran