El dl 20 de 02 de 2012 a les 15:46 +0000, en/na Jimmy O'Regan va
escriure:
> On 20 February 2012 15:26, Francis Tyers <[email protected]> wrote:
> > El dl 20 de 02 de 2012 a les 15:28 +0000, en/na Jimmy O'Regan va
> >> It might reduce the number of situations where the undefined behaviour
> >> arises, but it won't fix it - it will still be undefined where
> >> agreement is required, and the candidates differ in gender or number,
> >> for example.
> >
> > Well, it should fix it because it won't be undefined where the agreement
> > is required. You will always (effectively) get:
> >
> > ^direbech<adj><mf><sp>/sans reproche<adj><mf><sp>$
> >
> > Parts from the second LU won't get put on the first LU. Although maybe
> > if the number of tags is different, but I don't think so.
> 
> Eh? Do you mean 'transfer should check if more than one translation
> has been returned, and if so, choose one'? Probably, but I wouldn't
> eliminate the current behaviour entirely, because it is useful for
> finding testvoc errors.

Well, that's what the -b mode to apertium-transfer does, the same
behaviour as the VM. If more than one translation is returned by the
lexical transfer, just go with the first. 

I wouldn't eliminate it either (in fact it is the default behaviour),
but the testvoc could quite easily written to check for '\/.*\/' from
lt-proc -b, maybe somehting like this for LR:

$ lt-expand apertium-br-fr.br-fr.dix | grep -v ':<:' | cut -f1 -d':' |
sed 's/^/^/g' | sed 's/$/$/g' | lt-proc -b br-fr.autobil.bin  | grep
'\/.*\/'

Fran


------------------------------------------------------------------------------
Try before you buy = See our experts in action!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-dev2
_______________________________________________
Apertium-stuff mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/apertium-stuff

Reply via email to