Feature classes are created at load time.  Just create a feature
function and keep the $n$ language models as member variables.

Kenneth

On 06/10/14 13:33, Lars Bungum wrote:
> Den 10. juni 2014 19:01, skrev Kenneth Heafield:
>> I'm not sure what the end goal is here, but decide whether you want to
>> call the feature API generically moses/LM/* or directly e.g.
>> http://kheafield.com/code/kenlm/developers/, SRILM docs, IRSTLM, docs,
>> ...
> What I want to do is to create a feature that will score according to
> either one of _n_ (say 5) LMs according to some criterion.
> 
> I want to avoid loading the LMs inside my feature as this will obviously
> take time, and want to load them all ahead of decoding, which is done
> for the one configured in moses.ini in a baseline system.
> 
> Which is why I'm trying to find out where, exactly, in the moses code
> the LM is loaded.  I think I found that, thanks to Hieu's post now, but
> I am not yet sure how to alter the code to suit my needs. So I try to
> make sure that it's not possible within the current framework to avoid
> unnecessary engineering.  It's kind of hard to follow the flow of the
> files, where the moses.ini is read, what happens to the settings, and
> what makes it invoke the Load() for one of the LMs.
> 
> //LB
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to