Post-Editing Efficiency And Quality Assessment: A Comparative Analysis Of Google Translate And DeepL

Abstract

Neural Machine Translation (NMT) has revolutionized the translation industry by improving fluency, grammatical accuracy, and contextual understanding. However, its impact on post-editing efficiency and translation quality for domain-specific texts remains underexplored. This study compares the translation performance of Google Translate and DeepL, focusing on medical and legal texts. Using a simulated experimental setup, the study evaluates translation outputs based on editing time, automated metrics (BLEU, TER), and error analysis. The findings reveal that DeepL consistently outperforms Google Translate, requiring less editing time (M = 12.3 minutes), achieving higher BLEU scores (M = 80.3), and generating fewer lexical and syntactic errors. These results highlight DeepL’s suitability for domain-specific workflows requiring precision and accuracy. The study emphasizes the importance of selecting the right NMT tools to enhance productivity and translation quality in professional contexts. Future research should explore real-world applications, include additional NMT tools, and address cultural and linguistic nuances to broaden the understanding of NMT performance.