Description
The paper presents a comparison of two most popular Machine Translation paradigms for translation between related languages. Two language pairs on three different translation platforms were observed in the experiment. One pair represents really very close languages (Czech and Slovak), the other pair are slightly less similar languages (Slovenian and Croatian). The comparison is performed by means of three MT systems, one for each pair representing rule-based approach, the other one representing statistical (same system for both language pairs) approach to the task. The results were manually evaluated by native speakers of the target languages (linguists and students). The paradigms were compared and some surprising results were found (namely Statistical Machine Translation (SMT) paradigm seems to be a better choice as long as enough corpora are available and enough effort is put into the making of the translation system although the results suggest that some rule-based approach was used by the SMT system).
Primary author
Dr
Jernej Vičič
(University of Primorska)
