Hi, Tell me if i'm wrong, but, I don't think this format is supported by SRILM toolkit because it is not EXACTLY in SLF format. Sure, it is similar format, but SLF format does not allow multiple values for the "a" field nor a "r" field.
Here is a link for SLF format. http://labrosa.ee.columbia.edu/doc/HTKBook21/node257.html In order to rebuild the word graph, you have to parse the graph and recreate the "a" field (which corresponds to the translation score, i.e. the sum of the weighted feature functions). Hope it helps Regards, Loïc 2009/8/23 Chris Dyer <[email protected]> > The output word graph is in HTK Standard Lattice Format (SLF). A > number of tools (for example, SRILM's lattice-tool) read word graphs > in this format and can do things with it. You ought to be able to > find the path in the lattice corresponding to the decoder's best path. > > Chris > > On Sun, Aug 23, 2009 at 4:12 PM, Mostafa > Aghajani<[email protected]> wrote: > > Hi, > > > > I have used moses to generate a word graph (with –owg parameter) and here > > are some lines of the result: > > > > > > > > J=0 S=0 E=164 a=-0.149143, -0.0213871, -0.106304, > > -0.0113856, 0.999896 l=-6.31889 r=-50, 0, > -0.855577, > > 0, 0, 0 w=. > > > > J=1 S=0 E=72 a=-4.65502, -4.93344, -5.12308, > -4.84343, > > 0.999896 l=-8.26024 r=-10, 0, -1.49066, 0, 0, > > 0 w='s > > > > J=2 S=0 E=172 a=-6.19471, -5.827, -6.24869, -5.79268, > > 0.999896 l=-6.13104 r=-50, 0, -0.52219, 0, 0, 0 > w=, > > > > J=3 S=0 E=171 a=-2.49257, -2.43084, -6.54021, > -6.29025, > > 0.999896 l=-8.62969 r=-50, 0, -0.747215, 0, 0, > > 0 w=; > > > > J=4 S=0 E=150 a=-2.85263, -2.84781, -2.99573, > -2.83321, > > 0.999896 l=-11.8768 r=-40, 0, -1.09861, 0, 0, > > 0 w=homes > > > > J=5 S=0 E=148 a=-0.693147, -0.878071, -0.660357, > > -0.753772, 0.999896 l=-11.8506 r=-40, 0, -1.97716, 0, 0, > > 0 w=house > > > > J=6 S=0 E=152 a=-3.3322, -3.3322, -4.09434, -4.21951, > > 0.999896 l=-12.3013 r=-40, 0, -1.60944, 0, 0, 0 > > w=household > > > > J=7 S=0 E=142 a=-1.09861, -1.09861, -4.04305, > -4.29046, > > 0.999896 l=-15.0424 r=-30, 0, -0.510826, 0, 0, > > 0 w=painted > > > > J=8 S=0 E=141 a=-1.60944, -2.19723, -4.04305, > -4.29046, > > 0.999896 l=-13.9805 r=-30, 0, -0.510826, 0, 0, > > 0 w=small-scale > > > > J=9 S=0 E=137 a=-1.99243, -2.16905, -2.94444, > -2.90417, > > 0.999896 l=-12.5811 r=-30, 0, -2.19723, 0, 0, > > 0 w=tiny > > > > J=10 S=0 E=80 a=-3.04452, -0.734794, -7.72577, > -5.4807, > > 0.999896 l=-10.0289 r=-10, 0, -0.20067, 0, 0, > > 0 w=" is > > > > > > > > My question is how can I rebuild the word graph using these data? > > > > For example in line 1, is S=0 means that this a start node and E=164 > means > > that it connects to a node with S=164? If this is the answer (my guess), > why > > I can’t create the moses output translation (this is a small house.) > using > > this word graph? > > > > > > > > Best regards, > > > > M.Aghajani > > > > _______________________________________________ > > 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 > -- --- Loïc BARRAULT Docteur en Informatique LIUM - Université du Maine Tél. +33/0 2 43 83 38 27 http://www-info.univ-lemans.fr/~barrault ---
_______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
