Cultural Adaptation of Menus Using Machine Translation

This technical report presents a novel approach to the translation of culture-specific items (CSIs) in restaurant menus, focusing on the challenges posed by machine translation systems. The authors introduce the ChineseMenuCSI dataset, which is the largest collection of Chinese-English menu entries, annotated to facilitate a fine-grained analysis of CSIs. The report outlines a new methodology for automatic identification of CSIs that integrates human translation theories into large language model (LLM) processes, resulting in improved translation accuracy. The study highlights the limitations of existing machine translation systems in capturing cultural nuances and proposes techniques that outperform current methods. The authors detail the dataset's structure, which includes 4,275 bilingual entries categorized into CSIs and Non-CSIs, and discuss the implications of their findings for enhancing translation quality. The report emphasizes the importance of cultural awareness in translation and presents a comprehensive evaluation of the proposed techniques against traditional methods.