Learning to Combine Machine Translation Systems
Evgeny Matusov, Gregor Leusch, Hermann Ney · The MIT Press eBooks · 2008
This chapter shows how outputs of multiple machine translation (MT) systems may be combined in order to improve translation quality. It presents a method that computes a consensus translation from the aligned MT hypotheses. The method utilizes the enhanced alignment procedure of Matusov et al. (2006), as well as its extensions. The proposed system combination approach was evaluated on well-established Chinese-to-English and Arabic-to-English large-vocabulary translation tasks. Significant improvements in translation quality were obtained when compared with the best individual MT system.