Exploring the potential of GPT-4 as an interactive transcreation assistant in game localisation: A case study on the translation of Pokémon names

Generative Pre-trained Transformer (GPT) models are beginning to assist humans in complex creative endeavours. Transcreation is one of such creative processes, which in the context of translation consists of linguistically and culturally adapting a text or a part of it, including visuals, with the goal of maximising positive reception from a specific audience. Game localisation constantly engages in the transcreation of both real references (realia) and fictional elements (irrealia). Irrealia, albeit imaginary, are essential for gamers’ successful immersion and, thus, need to be properly adapted. This paper examines the potential of GPT models as assisting tools during game localisation processes. This exploratory study exemplifies how GPT-4 may be able to generate new language-specific irrealia via prompting to adapt fictional neologisms, focusing on pokémonikers, portmanteaus from the Pokémon game saga. It also describes how transcreation-related outputs can be used by translators as a pool of localisation options, for brainstorming, practice, and other purposes. The paper finds that the implementation of GPT chatbot-human interactions shows growing potential to support translators while adapting fictional elements, but also emphasises the current limitations of using GPT models, including accuracy, context-awareness, cultural sensitivity, and appropriateness, as well as potential ethical concerns.