Abstract (eng)
The present paper focuses on the comparison of the time translators need to post-edit a neuronally generated technical text and to translate a technical text without using machine translation. Both technical texts are analyzed in the language pair German-Polish. In addition, the translators' attitude to neuronal machine translation and its post-editing is examined. The hypotheses and research questions were based on the current theories.
First, an overview of the machine translation with a focus on neuronal machine translation and the attitude of the translators towards machine translation based on current research results is provided. Furthermore, the term post-editing and the attitude of translators to post-editing will be analyzed. The importance will be paid to the differences between post-editing and human translation, as well as to current research on the time required for post-editing and human translation. Afterwards, the state of the Polish language in the digital language industry will be discussed.
The empirical part presents the results of the study. Using post-edited and manually translated technical texts by 15 “advanced” students, called translators, together with screen recording videos, both the time of post-editing and translation without the use of machine translation and the attitude towards neural machine translation and its post-editing will be measured.
The results show that the post-editing of the neuronally machine-translated technical text in Polish takes less time than the manual translation from German into Polish. It has also been shown that machine translation makes the work of translators easier. However, the shorter time of post-editing depends on a wider range of factors than just a quick translation. The results are relevant to this paper and may contribute to the further improvement or development of neuronal machine translation.