Views on machine translation: An analysis on the performance of DeepL and Google Translate

Abstract

This paper comparatively evaluates the performance of two machine translation systems, namely DeepL and Google Translator. It starts with a brief overview of machine translation using neural networks, followed by a reflection concerning the performance of machine translation output using human and automatic evaluation. After that, this study addresses the difference in performance between both systems and identifies possible linguistic problems, possibly generated from these machine translation output. To carry on such comparison, this study employs an excerpt from Machado de Assis's Dom Casmurro, taken form the chapter "Olhos de Ressaca". This investigation uses the automatically generated translation outputs to critically analyze which system conveys more quality in their translation output and to comparatively examine linguistic errors generated by DeepL e Google Translate results. The evaluation criteria include the accuracy of the translation, the ability to maintain the meaning and structure of the original sentence, as well as fluency and adequacy of the resulting translation.