Linguistic evaluation of the first English-Lithuanian machine translation system

The paper deals with the most common problems of the machine translation (MT) process and the errors of the first English-Lithuanian MT prototype. These problems arose during the evaluation process of English-Lithuanian rule-based MT system which was the part of the project "Web Information Translation Facility" (2005-2007). We aim to test English-Lithuanian MT system’s output and identify different kinds of linguistic inaccuracies. MT system’s linguistic inaccuracies can be analyzed within different levels: lexicology, grammar, pragmatics, punctuation and spelling. Typological differences of English (an analytical language) and Lithuanian (a synthetical language) cause various errors and inaccuracies that appear during the process of translation. The paper overviews quality of the MT system and proposes the ways for its improvement