This thesis describes structural differences and translation asymmetries between English and Japanese. It describes possible ways to treat those phenomena in the framework of linguistics-based machine translation. As an application, the adaptibility of the multilingual linguistics-based MT system Its-2 to English-Japanese translation is tested, describing the enhancement of the system and its lexical databases, and the positive effect of using statistically acquired linguistic data. The translation quality of the English-Japanese version of Its-2 is finally compared to these of 2 other MT systems, on a set of 5 chosen sentences, showing strength on some points and weeknesses on others. It is also shown that hierarchical statistical post-editing can improve the overall translation quality of the system for daily language translation when the training corpus is not too small.