Abstract

Progression in AI translation has advanced greatly in the past couple of years, and with companies choosing to use it to translate foreign media, the subject of its use has become a topic of discussion. As a result, this research seeks to discover whether the current state of AI translation (DeepL) can successfully translate fictional dialogue from Japanese to English to the same standard a human translation is expected to have, as well as if the public can identify AI translation from human translation. The media will be analysed based on criteria gathered from previous research: Accuracy, Fluency, Appropriateness, and Expressionism. The public research will take place on a survey where participants are shown both translations side by side and asked which they think is the AI. The findings showed that DeepL does not meet the standard of human translation and can be easily identified by the public when put against a human translation. It not only translates literally and frequently mistranslates cultural and emotional expressions, but the inconsistent names and failure to adhere to the genres and theming of a text prove it inadequate, and in turn allow it to be identifiable in comparison to the human. The translation industry is better to use AI as an assistive tool rather than a singular tool. This research can serve as a basis for a larger exploration of AI translations ability or assist in other’s research.