Artimų kalbų mašininis vertimas
Mechanical Translation between Contiguous Languages

Summary/Abstract: The results of machine translation as one of the biggest challenges of today’s computational linguistics depend on many various criteria such as domain specificity and source and target language similarity. Recent projects have shown that machine translation among related languages (e.g., Slavonic) can be performed without a full-fledged analysis; good results can be achieved by analyzing only simpler constituents. One of such projects is the system Česílko developed at Charles University in Prague at the Institute of Formal and Applied Linguistics that has been extended from Slavonic languages to another language family, the Baltic languages. This paper presents the architecture of this system adapted for Lithuanian, describes used partial parser, explains similarities and differences between Czech (source language) and Lithuanian and the structure of the translation dictionary. Moreover the authors explain how they have evaluated the quality of translation.

• Details
• Contents