Statistical vs. Neural Machine Translation: A Comparison of MTH and DeepL at Swiss Post's Language Service
Lise Volkart, Pierrette Bouillon, Sabrina Girletti · Archive ouverte UNIGE (University of Geneva) · 2018
This paper presents a study conducted in collaboration with Swiss Post's Language Service that aims to compare the performance of a generic neural machine translation system (DeepL) and a customised statistical machine translation system (Microsoft Translator Hub, MTH) in terms of post-editing effort and quality of the final translation for the language direction German-to-French. The results for automatic and human evaluations show that DeepL is overall better than MTH, but its quality is underestimated by the BLEU score.