| Abstract [eng] |
The relevance of the topic is based on the fact that traditional food is linked to means of cultural preservation and many travellers seek to experience culture through gastronomy. An inaccurate menu translation negatively affects the reputation of the restaurant and the overall experience of the customers. In terms of novelty, the distinctiveness of culture-specific items (CSIs) has been addressed by many scholars, such as Davies (2003), Pedersen (2011) and Petrulionė (2012), yet there has been little research on Lithuanian restaurant translations and no research regarding the CSI translation accuracy generated by AI menu translation tools. The object of the research is culture-specific items in traditional Lithuanian cuisine restaurant menus, translated from Lithuanian into English. The aim is to analyse the most common issues of culture-specific item translation in traditional Lithuanian restaurant menus. In order to achieve the aim of the research, the following objectives have been set: 1. To provide an overview of the translation industry in relation to technologies, CSI characteristics and food discourse. 2. To analyse and classify the human translated CSIs of five traditional Lithuanian restaurant menus in terms of translator’s cultural knowledge, substitution accuracy, and strategy choice. 3. To analyse and classify the CSI translation accuracy generated by three AI menu translation tools. 4. To conduct a comparative analysis of human translated CSIs and AI translated CSIs concerning translation accuracy and strategy choice. The research methods employed in this study are qualitative and quantitative methods as well as the comparative approach. The research consists of an introduction, theoretical part with an emphasis on technologies and the translation peculiarities regarding CSIs and food related discourse. Moreover, there is a methodological section, in which the implemented methods are detailed. The research’s empirical part classifies and examines the accuracy of human and AI generated translation of CSIs. The research concludes with the findings obtained from the theoretical and empirical parts of the study. The research focuses on the translation of 30 CSIs that were referenced 154 times in the English versions of 5 traditional Lithuanian restaurants such as “Bernelių Užeiga”, “Etno Dvaras”, “HBH Palanga”, “Agotos Gryčia” and “Katpedėlė”. The translation variations were classified following a taxonomy proposed by Jan Pedersen (2011) and their accuracy was evaluated based on various traditional food related literature. The AI generated results were obtained by uploading an image of a synthetic menu to “MenuGuide” (2025), “MenuTranslator App” (2024) and “Menuly” (2023) mobile applications. The analysis revealed that cultural substitution and generalization were the most prominently applied strategies in both human and AI translations. The human translated restaurant menu analysis revealed that there were 154 translations and 101 errors, 33 of which were related to inaccurate cultural substitution. Other common errors were underspecification and translation inconsistency. Moreover, meat dish CSI translation was exceptionally problematic since 39 errors were found in 42 translation variations. The AI generated CSI translation analysis revealed that there were 118 translations and 135 errors. The most common errors were underspecification, inaccurate cultural substitution and inaccurate illustrations. Notably, there were 11 cases when AI generated completely unrelated translations. “MenuTranslator App” displayed the highest level of CSI translation accuracy, while “MenuGuide” was the least accurate. It is evident that human translation is more reliable than AI generated translation regarding the 30 CSIs. |
