Hi,

you are using the right setup (assuming that you want to
run the evaluation on both of the test sets now).

EMS is able to recognize that you ran a needed step
previously with the same tools on the same data sets
before. So, something must have changed in your new
experiments, maybe you updated the tools (EMS looks
at time stamps of tools such as the tokenizer).

With the verbose option you can see more details why
prior steps are invalidated, although admittedly the
displayed information is not necessarily easy to understand.

-phi

On Sun, Dec 4, 2011 at 11:13 PM, Dominikus Wetzel <[email protected]>wrote:

> Dear all,
>
> I am using EMS for my Moses experiments and have a successfully run
> experiment. Now, I would like to have additional evaluations based on
> different source and reference data. My thought was to just add new
> evaluation sections to the configuration file, however, when running EMS
> (without -exec and both with and without -continue), it suggests to
> rerun the entire experiment instead of just the new evaluations.
>
>
> Here are the relevant parts from the configuration file ("test" is from
> the original experiment, "test-new" should be the new evaluation):
>
> [EVALUATION:test]
> input-sgm = $data/test.$input-extension.sgm-src
> reference-sgm = $data/test.$output-extension.sgm-ref
> wrapping-frame = $input-sgm
>
> [EVALUATION:test-new]
> input-sgm = $data/test-new.$input-extension.sgm-src
> reference-sgm = $data/test-new.$output-extension.sgm-ref
> wrapping-frame = $input-sgm
>
>
> Can you please let me whether this is possible with EMS and if yes,
> guide me towards the correct configuration?
>
> Thanks a lot in advance.
> Best, Dominikus
> _______________________________________________
> 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