AI IN NONLITERAL LANGUAGE TRANSLATION: TRANSLATING MACEDONIAN PROVERBS AND IDIOMS
DOI:
https://doi.org/10.20544/teacher.29.07Keywords:
Artificial Intelligence, Machine Translation, Nonliteral Language, Proverbs, Idioms, AI TranslationAbstract
Translating proverbs and idioms poses a significant challenge for translators worldwide. Thus, it is no wonder that AI models seem to struggle, too. These expressions often rely on cultural context, figurative meaning, and linguistic creativity – things that AI models cannot really replicate. Recognizing this, this study examines how three AI models—ChatGPT, Gemini, and DeepSeek—handle the translation of Macedonian proverbs and idioms into English. The findings reveal several patterns: Explanation accuracy varies though all AI models consistently provide a two-part translation; they all favor direct translations over idiomatic equivalents, provide uniform translations of simpler proverbs, but show significant variability in word choice for more complex ones, and fail to maintain stylistic elements such as rhyme, alliteration, and wordplay.