From: <re...@us...> - 2007-09-27 05:03:34
|
Revision: 1463 http://mosesdecoder.svn.sourceforge.net/mosesdecoder/?rev=1463&view=rev Author: redpony Date: 2007-09-26 22:03:06 -0700 (Wed, 26 Sep 2007) Log Message: ----------- fixed a subtle bug in the word lattice decoder. it now acts sane. This experience also gave me more evidence for thinking that numbering the words in a phrase is not good-- we should be numbering the spaces between them. this would fix oh so many nasty things we've got (like the fact that we represent word drops with [n,n-1] etc). Modified Paths: -------------- trunk/moses/src/ConfusionNet.cpp trunk/moses/src/FloydWarshall.cpp trunk/moses/src/Hypothesis.cpp trunk/moses/src/InputType.cpp trunk/moses/src/InputType.h trunk/moses/src/Manager.cpp trunk/moses/src/WordLattice.cpp trunk/moses/src/WordLattice.h trunk/regression-testing/tests/lattice-distortion/to-translate.txt trunk/regression-testing/tests/lattice-distortion/truth/results.txt This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site. |