Cultural Adaptation of Menus: A Fine-Grained Approach
Zhonghe Zhang, Xiaoyu He, Vivek Iyer, Alexandra Birch
TL;DR
This work tackles the challenge of translating culture-specific items in restaurant menus by introducing the ChineseMenuCSI dataset (4,275 bilingual entries from UK Chinese menus) and a fine-grained three-category CSI taxonomy. It develops a Combined CSI Identification framework using RTT, CU, and HS to detect CSIs without parallel corpora, and proposes recipe-based and translation-theory–inspired prompting to improve CSI translations. Empirical results show LLMs outperform NMT on CSIs, with COMET gains up to 7+ points when external knowledge (recipes) and translation strategies are used, while NMT remains better for non CSIs. The study demonstrates that integrating translation theory into prompts and leveraging external culinary knowledge substantially enhances cross-cultural translation quality and offers a scalable framework for culture-aware MT.
Abstract
Machine Translation of Culture-Specific Items (CSIs) poses significant challenges. Recent work on CSI translation has shown some success using Large Language Models (LLMs) to adapt to different languages and cultures; however, a deeper analysis is needed to examine the benefits and pitfalls of each method. In this paper, we introduce the ChineseMenuCSI dataset, the largest for Chinese-English menu corpora, annotated with CSI vs Non-CSI labels and a fine-grained test set. We define three levels of CSI figurativeness for a more nuanced analysis and develop a novel methodology for automatic CSI identification, which outperforms GPT-based prompts in most categories. Importantly, we are the first to integrate human translation theories into LLM-driven translation processes, significantly improving translation accuracy, with COMET scores increasing by up to 7 points.