Discontinuous Statistical Machine Translation with Target-Side Dependency Syntax

For several languages only potentially non-projective dependency parses are readily available. Projectivizing the parses and utilizing them in syntax-based translation systems often yields particularly bad translation results indicating that those translation models cannot properly utilize such information. We demonstrate that our system based on multi bottom-up tree transducers, which can natively handle discontinuities, can avoid the large translation quality deterioration, achieve the best performance of all classical syntax-based translation systems, and close the gap to phrase-based and hierarchical systems that do not utilize syntax.