Comparing Neural Machine Translation and Human Translation of Dickinson’s Because I Could Not Stop for Death

Abstract

The development of neural machine translation (NMT) as a branch of artificial intelligence has significantly influenced translation practices, including literary translation. However, the capacity of NMT systems to preserve the aesthetic and stylistic dimensions of poetry remains contested. This study compares the performance of DeepL and a human translation in rendering Emily Dickinson’s poem Because I Could Not Stop for Death into Indonesian. The study employs a qualitative line-by-line comparative analysis focusing on equivalence, diction, shifts, and stylistic effects. The machine translation was generated through DeepL using default settings without post-editing. The findings reveal that DeepL demonstrates strong lexical accuracy and structural consistency, but tends to produce literal renderings that reduce metaphorical depth and aesthetic nuance. In contrast, the human translation shows greater flexibility in handling figurative language, rhythm, and emotional tone, resulting in a more poetic and communicative target text. These findings indicate that while DeepL can function as an initial assistive tool in poetry translation, human interpretive intervention remains essential to achieve artistic and stylistic adequacy.