Evaluating accuracy and fluency in machine and human translation: a genre-based multidimensional quality metrics (MQM) study in an English as a foreign language (EFL) context

Despite the growing use of Machine Translation (MT) in translation practice, limited studies have examined genre-based quality comparison between machine and Human Translation (HT) using Multidimensional Quality Metrics (MQM) framework, with limited evidence on accuracy and fluency. This study aims to compare the quality of Machine Translation (MT) output generated by Google Translate with Human Translation (HT) produced by undergraduate EFL student translators rather than professional translators in English-to-Indonesian translation across three genres informative, expressive, and operative texts focusing on two MQM dimensions: accuracy, covering mistranslation, omission, addition, and terminology errors; and fluency, including grammar, word-choice, cohesion, and readability issues. Accuracy and fluency were selected because they represent core MQM dimensions relevant to evaluating whether translated texts successfully preserve source-text meaning while remaining grammatically acceptable, coherent, and readable in the target language. Employing a qualitative comparative text-analysis design, the English to Indonesian translation of MT output generated by Google Translate and HT produced by EFL students are compared. The findings indicate that Google Translate performs relatively in translating informative texts, particularly in transferring factual content and maintaining basic lexical accuracy. In contrast, EFL student translation demonstrates stronger ability to interpret contextual meaning, preserve genre-specific style, adjust register, and maintain communication.