Morphosyntactic Target Language Matching in Statistical Machine Translation

While the intuition that morphological preprocessing of languages in various applications can be beneficial appears to be often true, especially in the case of morphologically richer languages, it is not always the case. Previous work on translation between Nordic languages, including the morphologically rich Finnish, found that morphological analysis and preprocessing actually led to a decrease in translation quality below that of the unprocessed baseline.
In this paper we investigate the proposition that the effect on translation quality depends on the kind of morphological preprocessing; and in particular that a specific kind of morphological preprocessing before translation could improve translation quality, a preprocessing that first transforms the source language to look more like the target, adapted from work on preprocessing via syntactically motivated reordering. We show that this is indeed the case in translating from Finnish, and that the results hold for different target languages and different morphologicalanalysers.