Abstract：Focused on the issue that lexicalized reordering models are faced with context independence and sparsity in machine translation, a reordering table reconstruction model based on semantic content for reordering orientation and probability prediction was proposed. Firstly, continuous distributed representation approach was employed to acquire the feature vectors of reordering rules. Secondly, Recurrent Neural Networks (RNN) was utilized to predict the reordering orientation and probability of each reordering rule that represented with dense vectors. Finally, the original reordering table was filtered and reconstructed with more reasonable reordering probability distributions for the purpose of improving the reordering information accuracy in reordering model as well as downscale the reordering table to speed up subsequent decoding process. Experimental results show that reordering table reconstruction model can provide BLEU point gains (+0.39) for Chinese to Uyghur translation task.
