Hi Daniel,

Thanks for the help.

In fact, I found that the option foe moses-chart to output those  
details is -T or -translation-detail. Also, I tried most of the  
options related to logging the process, but couldn't get them work as  
they are described in --help. May be I didn't do it correctly.

Anyway, I also could finally decode that log which confirms what you said:

Trans Opt 8 : seems to be the sentence identifier rather than  
translation opt(ion)
[1..8] : the source span covered by this rule
[8..8]=.   [7..7]=parameter   [3..6]=X   [2..2]=is [1..1]=this :  
granulated source side spans matched to the source right-hand-side of  
the applied rule
X ->dies ist X @-@ parameter . : the target expansion rule of the applied rule
2-2 : alignment between variables in the right hand side of the rule  
and source pC=-0.499665, c=-0.392943 -1.92099<<-3.90865, 0, -24.6762,  
-2.35267, -5.79759, -2.72686, -10.9766, 2.99969, 0>> : costs and scores

This interpretation welcomes comments as well.

Thanks and regards,
Rasul.

Quoting Daniel Schaut <[email protected]>:

> Hi Rasul,
>
> More information about the trace option can be found here:
> http://www.statmt.org/moses/?n=Moses.SyntaxTutorial#ntoc5
>
> But you're right the output probably varies depending on your model.
> Consider the following SL sentence containing eight tokens:
> this is an ' out ' parameter .
>
> A random line of my -t flag's output looks like the following (hierarchical
> phrase-based):
> Trans Opt 8 [1..8]: [8..8]=.   [7..7]=parameter   [3..6]=X   [2..2]=is
> [1..1]=this  : X ->dies ist X @-@ parameter . :2-2 : pC=-0.499665,
> c=-0.392943 -1.92099<<-3.90865, 0, -24.6762, -2.35267, -5.79759, -2.72686,
> -10.9766, 2.99969, 0>>
>
> Trans Opt 8 [1-8]: = current no. of hypothesis (identifier) + total no. of
> spans covered by hypothesis, as defined by token position,
> e.g. [7..7]=parameter = SL span, as defined by token position,
> or [3..6]=X = glue rule covered by glue grammar
> : = delimiter for SL and TL spans,
> X -> = rule used to generate TL span,
> dies ist ein X @-@ parameter . = generated TL span,
> 2-2 = ???,
> ...
>
> Note that these are just educated guesses derived from the above-mentioned
> link. Feel free to correct or comment.
>
> Regards,
> Daniel
>
> -----Ursprüngliche Nachricht-----
> Von: [email protected] [mailto:[email protected]] Im
> Auftrag von [email protected]
> Gesendet: 28 May 2012 12:13
> An: [email protected]
> Betreff: [Moses-support] Translation details log for chart decoder.
>
> Hi all,
>
> I'm trying to read the output of chart decoder for -translation-details (T)
> option and also -report-segmentation.
> However, none of the outputs look like what have been described by
> phrase-based and syntax-based tutorials on the website. I was wondering if
> anybody has any idea of where such information could be found, as it was
> difficult to interpret intuitively. Thanks in advance.
>
> Best regards,
> Rasul.
> _______________________________________________
> 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