Abstract (eng)
Despite the rapid development of machine translation, recent studies (e. g. Abdulaal 2022; Bolognesi & Horvat 2022; Zajdel 2022) have shown that it has not reached the level of human translation competence, especially with regard to literary translation. Since a comprehensive study on the general quality of literary machine translation would not be feasible, this thesis deals with only one aspect characteristic to literature: the metaphor. It focuses on the question of whether machines are currently capable of successfully translating figurative language in a novel – in this case David Copperfield by Charles Dickens (1850) – and which translation procedures they typically use. For this purpose, 38 metaphors were extracted from the novel and translated by the MT systems Google Translate and DeepL; a German translation of the novel by Gustav Meyrink (1910) served as a comparative text. To conduct the study, the source text metaphors were categorized according to their length and analyzed in terms of the translation methods applied. The analysis revealed that the MT systems translated the metaphors predominantly literally and used merely a few other procedures besides word-for-word-reproduction, including the transformation of the metaphor into non-metaphorical expressions. In contrast, the human translator made use of a greater variety of translation procedures. Moreover, Google Translate generally proved unsuitable for the transfer of figurative language, given the large number of erroneous translations, while DeepL frequently produced idiomatic translations. Nevertheless, considering the overall lack of competence of the MT systems to produce target texts equivalent in function and effect as shown in the present paper, MT cannot yet be used for the translation of metaphors in literature without considerable post-editing effort.