Machine translation (MT) has been part of the translation industry for many years, but the emergence of large language models (LLMs) has truly turned the field upside down. Unlike earlier translation systems, LLMs such as ChatGPT can generate fluent and stylistically adaptable translations. Combined with tools like Google Translate and DeepL, the work of translators and journalists has changed, as MT is increasingly integrated into their daily practice. Today, MT is used in a wide range of contexts; from casual communication to professional, medical, and legal settings.
However, the translation of metaphors remains a major challenge. Metaphors are figurative by nature and often culturally specific, making them vulnerable to amusing but also potentially serious mistranslations.  Although MT quality has improved greatly in the past years, even the most advanced systems still make mistakes or invent information. At present, no systematic research identifies which metaphors become problematic in MT and how mistranslations affect understanding.
The Machine-Translation Metaphor (MTM) project investigates which metaphors cause problems in MT, why these issues arise and under which circumstances, and how they affect users and society. It examines metaphors as simultaneously linguistic, cognitive, and cultural phenomena, and studies MT from three perspectives: the Machine, the Professional User, and the Non-expert Reader. The PhD project focuses on the second perspective, exploring how professional users, i.e. literary translators and journalists, revise and reflect on machine-translated metaphors and how this shapes their practice.

• large language models
• machine translation
• metaphor studies
• the human digital world

Share on Facebook Share by Bluesky Share on LinkedIn Share by WhatsApp Share by Mastodon