engleski

Fine-grained human evaluation of NMT applied to literary text: case study of a French-to-Croatian translation

Even though neural machine translation (NMT) has demonstrated phenomenal results and has shown to be more successful than previous MT systems, there is not a large number of works dealing with its application to literary text. This results from the fact that literary texts are deemed to be more complex than others because they involve more specific elements such as idiomatic expressions, metaphor, a specific author’s style, etc. Regardless of this fact, there is a growing body of research dealing with NMT applied to literary texts, and this case study is one of them. The goal of the present paper is to conduct an in- depth, fine-grained evaluation of a novel translated by Google Translate (GT) in order to reach detailed insights into NMT performance on literary text. In addition, the paper aims to include for the first time, to the best of our knowledge, the French-Croatian language combination.

neural machine translation ; literary text ; French ; Croatian

nije evidentirano