Hi Tino, would it be possible to do this with CG or would this need to
implemented in some new program? Anyhow, I suppose this involves a lot
of work writing clever rules.

My rational for target language smoothing is:
a) my bad experience of the performance of the old pair sv-da with lots
   of blatant errors made by the tagger.
b) target language smoothing is very easy to implement.

When doing lemmatisation the tagger isn't used, as far as I can see.
This leaves all inherent ambiguities in the Swedish dictionary to be
handled somehow.

So, I just proposed a quick fix that might improve the result
significantly in this special case (and maybe other cases with a lot of
ambiguity that is not properly handled). Other means my be more adequate
and/or more effective.

Yours, Per Tunedal

On Fri, Mar 4, 2016, at 08:53, Tino Didriksen wrote:
> On 4 March 2016 at 07:52, Per Tunedal
> <[email protected]> wrote:
>> Yes, of course! That has always seemed a bit unnatural to me. It's
>>
harder to decide on the right source language lemma before translating
>>
than doing it after translation.
>
> I almost entirely disagree, and I've got experience and data to back
> it up. Target language smoothing does not help much, if your source
> language analysis is good.
>
> You can disambiguate the source language to nigh-100% if you use more
> analysis levels, such as dependency and semantics. This is what we do
> at GrammarSoft / VISL. It works.
>
> Apertium could also do it this way, and it would benefit all languages
> built from a specific source.
>
> -- Tino Didriksen
> ----------------------------------------------------------------------
> --------
> Site24x7 APM Insight: Get Deep Visibility into Application Performance
> APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month
> Monitor end-to-end web transactions and take corrective actions now
> Troubleshoot faster and improve end-user experience. Signup Now!
> http://pubads.g.doubleclick.net/gampad/clk?id=272487151&iu=/4140
> _________________________________________________
> Apertium-stuff mailing list [email protected]
> https://lists.sourceforge.net/lists/listinfo/apertium-stuff
------------------------------------------------------------------------------
_______________________________________________
Apertium-stuff mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/apertium-stuff

Reply via email to