On Mon, Dec 17, 2012 at 08:24:38PM +0000, Francis Tyers wrote:
> El dl 17 de 12 de 2012 a les 23:18 +0100, en/na [email protected] va
> escriure:
> > On Mon, Dec 17, 2012 at 07:30:39PM +0100, Tino Didriksen wrote:
> > > To me, those examples have nothing to do with domain or style. They can be
> > > solved in better general ways with a mix of CG, semantics, and lexical
> > > selection. And I know this, 'cause it's what our non-Apertium
> > > closed-source
> > > translation engine uses.
> > >
> > > For example, the English text
> > > "The key to the piano will let you unlock it and hit the keys of the
> > > piano which are in harmonic key."
> > > translates into Danish as
> > > "Nøglen til klaveret vil lade dig låse det op og trykke på tasterne af
> > > klaveret som er i harmonisk toneart."
> > >
> > > In this example, the differences are resolved during lexical selection.
> > >
> > > We have a total of 2 domains, for cases where the same word in the same
> > > context truly means something entirely different. Introducing many tiny
> > > domains is counterproductive when you could instead solve it in a general
> > > way that will benefit all domains.
> > >
> > > Don't lock away good translations in domains when it's clearly not needed.
> >
> > I think I agree. You have 3 different meanings of "key" here.
> > How did you differentiate between the 3 different meenings of the homonym?
>
> With rules. I would do: default translation "nøgle", context
> translations "harmonic key" -> "harmonisk toneart" and "hit * key" ->
> "trykke * taste" (or something like that.)
>
> > Whet is your "non-Apertium closed-source translation engine"?
> > A description?
>
> GramTrans http://visl.sdu.dk/~eckhard/pdf/MTsummit07_final.pdf
I read this paper, and what I understood is that GramTrans employs a number of
MT techniques, including the ones we use in Apertium. But they also use the
concept relations that I am proposing, and some statistical MT, and more.
I found that their system looked quite advanced. Could it be considered
state of the art?
If so, would we want to also use some of the techniques they use?
It seems that their system basically is rule-based, on top of some
grammar analysis. I would think that this would need an architecture
in Apertium that is quite modularized ("The Unix way").
Myself not being familiar with the code of Apertium at all, is this so?
And could a module with use of concept reletions be easily included in the stack
of translation modules?
How much are we doing of what GramTrans is doing and are there plans to go
further that way?
best regards
Keld
------------------------------------------------------------------------------
LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial
Remotely access PCs and mobile devices and provide instant support
Improve your efficiency, and focus on delivering more value-add services
Discover what IT Professionals Know. Rescue delivers
http://p.sf.net/sfu/logmein_12329d2d
_______________________________________________
Apertium-stuff mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/apertium-stuff
