Forest-to-String Statistical Translation Rules

In this paper, we propose forest-to-string rules to enhance the expressive power of tree-to-string translation models. A forestto-string rule is capable of capturing nonsyntactic phrase pairs by describing the correspondence between multiple parse trees and one string. To integrate these rules into tree-to-string translation models, auxiliary rules are introduced to provide a generalization level. Experimental results show that, on the NIST 2005 Chinese-English test set, the tree-to-string model augmented with forest-to-string rules achieves a relative improvement of 4.3% in terms of BLEU score over the original model which allows treeto-string rules only.