Word re-ordering and dynamic programming based search algorithm for statistical machine translation

In this work, a new search procedure for statistical machine translation (SMT) is proposed that is based on dynamic programming (DP). The starting point is a DP solution to the traveling salesman problem that works by jointly processing tours that visit the same subset of cities. For SMT, the cities correspond to source sentence positions to be translated. Imposing restrictions on the order in which the source positions are translated yields a DP algorithm for carrying out the word re-ordering in SMT efficiently. A simple data-driven search organization allows the algorithm to prune unlikely translation hypotheses. Search restrictions especially useful for the translation directions German-to-English and English-to-German are presented. A generalization of these re-ordering restrictions is given that is applicable to several different translation directions. Translation results are reported with a widely used SMT model.