El dt 30 de 04 de 2013 a les 09:19 +0200, en/na Kevin Brubeck Unhammer
va escriure:
> Jonas Fromseier Mortensen
> <[email protected]> writes:
> 
> >     Supporting nn→da as well should present only minor additions,
> >     mostly to
> >     your make system. You'd have two modes generated, nn→da and nb→da,
> >     which
> >     would share everything that comes _after_ bidix (ie. structural
> >     transfer
> >     and da generation). On the nn/nb side of the pipeline, you could
> >     probably snatch the monodixes, prob files and CG's from
> >     apertium-nn-nb
> >     without changes. 
> >     
> >     The only remaining thing is bidix. Here I would keep one master
> >     bidix
> >     from no (nn+nb) to da, which is processed by an XSLT script into
> >     two
> >     different bidixes before compilation. Most entries would be the
> >     same,
> >     but some would be marked nn-only or nb-only. This kind of thing
> >     happens
> >     in a lot of apertium pairs, and should be no trouble to set up.
> >     
> >
> > Hi Kevin 
> > Thanks for the advice! It sounds very doable the way you put it. What
> > do you think would be the best way to go about it? make a set of
> > transfer rules for nb-da first, cg for the same and then work on
> > adding nn support after? I'm sure there's lots of quirks and idiomatic
> > expressions that my nb-da t1x wouldn't catch with nn as sl.
> 
> You'll find lots of nn-typical syntax in nb text and the other way
> around, so I'm actually pretty confident one t1x would cover both. You
> could of course initially focus your testing on nb.
> 
> CG for both nn and nb should be copied from apertium-nn-nb, so no need
> to start something new there.
> 
> A CG for da would be great, but would only be useful for the other
> direction. It could probably be based on the nb CG, deleting stuff that
> doesn't work before adding new stuff (actually, I would first run a big
> corpus through the CG with --trace and delete any unused rules, to make
> it easier to deal with). But if you work on no→da first, the da CG would
> not be useful yet.

On the other hand, a Danish CG would be useful for other pairs too, like
sv-da :)

I really think that given the data we have available, it's worth going
for a bidirectional pair.

F.


------------------------------------------------------------------------------
Introducing AppDynamics Lite, a free troubleshooting tool for Java/.NET
Get 100% visibility into your production application - at no cost.
Code-level diagnostics for performance bottlenecks with <2% overhead
Download for free and get started troubleshooting in minutes.
http://p.sf.net/sfu/appdyn_d2d_ap1
_______________________________________________
Apertium-stuff mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/apertium-stuff

Reply via email to