Sentence-structure errors of machine translation into Slovak

Abstract

The study identifies, classifies and analyses errors in machine translation (MT) outputs of journalistic texts from English into Slovak, using error analysis. The research results presented in the study are pioneering, since the issue of machine translation – with its strong interdisciplinary character and novelty – has not yet been studied in the Slovak academic environment. The evaluation of the errors is based on a framework for classification of MT errors devised by Vaňko, which was arranged for the Slovak language. The study discusses and explains the issues of sentence structure, including predicativeness, syntactic-semantic correlativeness, and a modal and communication sentence framework. We discovered that the majority of the errors are related to the categories of agreement, word order and nominal morpho-syntax. This fact clearly correlates with features of journalistic texts, in which nominal structures and nouns in all realizations are used to a great extent. Moreover, there are some serious differences between the languages which limit and affect the quality of translation.