The perception of Multimedia Translation students on automated translation versus human translation for localising a website

Abstract

This study describes a pre-experimental and exploratory design developed over two academic years, in which 57 students from a higher education institution were faced with the simulation of a reverse translation project (Spanish-English) for the website of a mock brewery located in the Andalusian city of Cordoba (Spain). The website was created ad hoc for this task and contained a large number of terms related to Andalusian cuisine, as well as cultural references, localisms and idiomatic expressions with a humorous undertone. After performing a creative translation to transfer the cultural and humorous features successfully, participants obtained versions generated with at least two automated translation tools (DeepL or SmartCat, and ChatGPT) to compare their own translations with those generated by machine translation (MT) and artificial intelligence (AI). Finally, they assessed the three target texts through a questionnaire. This paper presents the results obtained from the activity, with the aim of providing an insight into the Multimedia Translation students' perception of the use of MT and AI, specifically regarding five key aspects of the translation process: the overall quality of the target text, terminological accuracy, linguistic coherence and fluency, interpretation of the socio-cultural context and communicative intention, and grammatical and syntactic accuracy. The study confirms that human translation (HT) is rated higher than MT and AI in all the categories, and that this gap is more pronounced when it comes to the subjective aspects of language. More specifically, HT stands out for its coherence, fluency and ability to interpret contextual and socio-cultural factors in communication.

Keywords
multimedia translation; machine translation; artificial intelligence; human translation; localisation