El ds 02 de 04 de 2011 a les 17:53 -0400, en/na Ahsan va escriure:
> Hi all,
> 
> 
> This is regarding the adopting of the existing ur-hi language pair.
> Currently, the ur-hi pair has much work done with both urdu and hindi
> languages. 
> 
> 
> == Motivation ==
> 
> 
> The field of machine translation has gained much traction in the past
> couple of decades. Its use of translating between different language
> pairs is of political, military and international importance.
> Apertium was quite interesting to me when I first came across it given
> that it has an excellent documentation. As a newbie to MT, it walked
> me through some basic concepts on its wiki which has provided me a
> good base to build upon. 
> 

Thanks for the compliment ! 

> 
> 
> == Why must my project be selected? ==
> 
> 
> Looking at the state of this language pair, it can be made stable
> considering that all the initial work is done. Knowing both the
> languages personally I would be able to comprehend the transfer rules
> between them more easily.


I think "must" here sounds a bit presumptuous.

> == Goals ==
> 
> 
> *Convert the words from the apertium_hi_WX.dix from WX notation to
> unicode encoding. Unicode encoding makes it consistent to use with the
> apertium engine. Also, make the tags in the IIIT morphologic analyzer
> to be along the same lines as the tags in apertium language pair.

This should be done before the application is in. E.g. the deadline is
the 8th April. We've had a lot of people try this, but even after weeks
they have not succeeded. And I can't quite work out why. If you are
familiar with Hindi and the tagset, it shouldn't take more than a day.

> 
> *Make the bidix more complete by adding more words to the bilingual
> dictionary. And, finally test it.
>  The bidix file is apertium-ur-hi.ur-hi.dix

How are you going to do this ?

> *There is presently no parts of speech tagger for this. This tagger
> will be added to  which will be the new file.
> apertium-ur-hi.ur-hi.tsx

This doesn't sound right. Or are you planning to make a single tagger
for both Hindi and Urdu ? This is feasible, but then you will need to
make sure that the tagsets are identical.

> 
> *Adding tranfer rules before target language tagger training. The
> rules need to be added to the file.
>   apertium-ur-hi.ur-hi.t1x.
> 
> 
> *Train parts of speech tagger for the target language by using three
> stage transfer command,
>  apertium-transfer
>  apertium-tagger-tl-trainer
> 
> 
> *Finally, quality control tests would be run. The three test to be
> carried out are:
> -corpus test
> -regression test
> -testvoc

Which corpus are you planning to use ? 

[...]

> This is not complete by any means. Any feedback to improve it will be
> much appreciated.

Thanks for your suggestions. I suggest you read also my comments to
n0nick (Sagie), and his project proposal too.

Fran



------------------------------------------------------------------------------
Create and publish websites with WebMatrix
Use the most popular FREE web apps or write code yourself; 
WebMatrix provides all the features you need to develop and 
publish your website. http://p.sf.net/sfu/ms-webmatrix-sf
_______________________________________________
Apertium-stuff mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/apertium-stuff

Reply via email to