<![CDATA[This study aims to compare the quality of machine translation between Google Translate (GL) and DeepL (DL) in translating Chairil Anwar's poetry from Indonesian to German. Although AI-based machine translation systems perform well in technical and informative texts. Literary translation, especially poetry, remains challenging because it involves language style, emotion, metaphor, and cultural nuances. Poetry, therefore, serves as a relevant object for testing the ability of Neural Machine Translation systems to capture poetic quality and implicit meaning. This study adopts a qualitative case study approach to examine the German translations of the poem “Aku” produced by GT and DL. The analysis applies Koponen’s framework of translation errors, with the human translation by Karwath as a reference for translation quality and an established literary analysis as a reference for source text meaning. The data are complemented by dialogic expert interviews, analyzed thematically and triangulated. The results show that both machines make all three categories of translation errors based on Koponen’s theory. GT translates partially and tends to choose rougher and bolder diction, while DL tends to smooth the language, which in some lines feels less powerful or rather flat. Overall, no machine translation is able to fully reproduce the meaning, feel, and emotional nuance of the source text in poetic translation. Human involvement is still necessary to refine the translation results. However, between the two, DL can be considered the better option because it avoids serious errors with homonyms that are not identified by GL.]]>
Copyrights © 2026
