STROJOVÝ PREKLAD ZALOŽENÝ NA NEURÓNOVÝCH SIEŤACH – SĽUBNÁ CESTA PREKLADU Z ANALYTICKÝCH JAZYKOV DO FLEKTÍVNEJ SLOVENČINY?

In this paper, we deal with the issues of machine translation based on neural network. We explain the fundamental principles of this approach to automatic translation. Based on the analysis of statistical and neural machine translations from English into Slovak, we compare the quality, and/or the qualitative shifts after changing the approach to machine translation, from statistical to neural networks. The results reveal that neural machine translation achieves better results in fluency and grammatical correctness of translation, but the representation of semantically inadequate translations increases in examined corpus of journalistic texts.