Hi John,

yes, the translation model is trained on case-sensitive data,
and the recaser only uppercases the first word of each sentence.
For some of the results, check out ACL WMT-08 workshop paper.
For other language pairs we so slight increases or drops in
scores (in the 0.1-0.3 range) compared to the traditional approach.

-phi

On Thu, Jul 17, 2008 at 9:17 PM, John D. Burger <[EMAIL PROTECTED]> wrote:
> Philipp Koehn wrote:
>
>> just to add one comment: We have recently experimented
>> with training models on truecased model (where only the
>> first word of each sentence is converted into its most common
>> casing), with mixed results. Such a truecaser should also
>> be adjusted to deal with ALL-CAPS HEADLINES and so on.
>
> Hi Philipp -
>
> Just to clarify, you mean you have built case-sensitive translation
> models, with no separate recaser?  Or are you talking about a
> slightly different approach to training a recaser?
>
> Can you expand on your "mixed results" comment?  Thanks!
>
> - John D. Burger
>   MITRE
>
> _______________________________________________
> Moses-support mailing list
> [email protected]
> http://mailman.mit.edu/mailman/listinfo/moses-support
>
>
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to