Real-world challenges in application of MT for localization: the Baltic case

In this paper we share our experience from implementing machine translation in localization into relatively small languages of the three Baltic countries – Latvian, Lithuanian, and Estonian. We describe our approach in improving terminology translation and consistency by preprocessing of the source text and performing term integration. We present results of a formal evaluation of MT impact on the productivity of translators’ work. Quality, usability, increase of productivity of translation process has been evaluated. Our evaluation shows that the use of the SMT suggestions in addition to the translation memories in the SDL Trados CAT tool leads to the increase of translation performance by up to 32.9% while maintaining an acceptable quality of translation.