James, The simple distortion model penalizes reordering (assuming its weight has the correct sign), but the lexicalized reordering model gets more at what you're suggesting, and pretty much always improves BLEU. Kevin
On Wed, Nov 6, 2013 at 12:08 PM, Read, James C <[email protected]> wrote: > Hi, > > I was just reading through the phrase tutorial on the Moses website and > ran into this: > > > Each of these models contributes information over one aspect of the > characteristics of a good translation: > > * The phrase translation table ensures that the English phrases and > the German phrases are good translations of each other. > * The language model ensures that the output is fluent English. > * The distortion model allows for reordering of the input sentence, > but at a cost: The more reordering, the more expensive is the translation. > * The word penalty ensures that the translations do not get too long > or too short. > > I'm not sure if I'm reading this right. The more reordering the more a > translation is penalised. Surely that's an over generalisation and there > exist a set of phrase pairs (particularly for language pairs of different > word order) that we would want to penalise if they were not reordered but > reward for fulfilling the statistically observed necessity for some measure > of reordering. > > James > > _______________________________________________ > 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
