Scholarlink

This study examines the translation of metaphorical expressions in economic discourse using large language models (LLMs), such as ChatGPT-5, Gemini, and Perplexity. Fifty German metaphors were translated into English and Slovak to compare the models’ performance in high- (English) and low-resource (Slovak) languages. Both qualitative and quantitative evaluations were employed to assess metaphorical equivalence and the quality of the translations overall. The results revealed significant differences between the models and language pairs, with proprietary systems demonstrating greater fluency and figurative accuracy. This study deepens our understanding of how LLMs handle figurative meaning and identifies areas for improvement in both high- and low-resource language contexts.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Bewertung von Sprachmodellen bei der Übersetzung metaphorischen Denkens in der Wirtschaftssprache

  • 章节

  • 出版模式：

    开放获取

  • 发表日期：

    2026年

查看全文
  • Filip Kalaš,
  • Pavol Lipták

摘要

This study examines the translation of metaphorical expressions in economic discourse using large language models (LLMs), such as ChatGPT-5, Gemini, and Perplexity. Fifty German metaphors were translated into English and Slovak to compare the models’ performance in high- (English) and low-resource (Slovak) languages. Both qualitative and quantitative evaluations were employed to assess metaphorical equivalence and the quality of the translations overall. The results revealed significant differences between the models and language pairs, with proprietary systems demonstrating greater fluency and figurative accuracy. This study deepens our understanding of how LLMs handle figurative meaning and identifies areas for improvement in both high- and low-resource language contexts.