On 15/06/2013 10:11, Per Tunedal wrote:
> Hi,
> I've tested several models fr-sv and found that syntax in source will
> yield a strange word for word translation of "je ne sais pas de qui tu
> parles ." to "jag inte vet inte om vem du pratar om ."
>
> I've ran the translation with the option -v 3 but cannot interpret the
> output, my log file ends with:
>
> worse than matching hyp 8778, recombining
> worse than matching hyp 8778, recombining
> 0 1 2 3 4 5 6 7 8 9 10
> 1 20 20 20 19 17 20 12 19 20 0
> 20 0 0 2 16 6 1 2 0 0
> 184 0 5 0 0 1 0 2 0
> 81 3 0 0 0 0 0 0
> 141 0 0 0 0 0 0
> 200 0 0 0 0 0
> 109 0 0 0 0
> 94 0 0 0
> 62 0 0
> 128 0
> 1
> BEST TRANSLATION: 8778 S </s> :0-0 : c=-0.037
> core=(-0.434,0.000,0.000,0.000,0.000,0.000,0.000,0.000,0.000) [0..10]
> 7725 [total=-0.464]
> core=(-5.212,0.000,-16.360,-8.939,-11.236,-10.248,-11.047,8.999,8.999)
> Best path: 7725 <= 6625 <= 5517 <= 4264 <= 3094 <= 1864 <= 640 <= 230 <=
> 181 <= 0 <= 1 <= 21 <= 41 <= 61 <= 84 <= 102 <= 121 <= 142 <= 161 <= 0
>
> How can I follow the search? How can I find the rules used?
to see the rules used to create the best output, add the argument
-T transOpt.txt
when running the decoder.
to follow the search is a good question, but i don't have an easy
answer. The best way to start is to translate small sentences and walk
thru the code, adding your debugging messages
>
> Yours,
> Per Tunedal
> _______________________________________________
> 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
