Translating English Number Idioms into Arabic: A Neural Machine Translation Evaluation

Abstract

Idiomatic expressions, particularly those involving numbers, pose significant challenges for neural machine translation (NMT) systems due to their non-literal meanings and cultural specificity. This study investigates the efficacy of NMT systems, specifically Google Translate and Chatgpt, in translating English number idioms into Arabic. A corpus of 30 English number idioms was investigated and their translations analyzed in terms of literalness, cultural appropriateness, and semantic accuracy. The findings reveal that while some idioms are accurately translated with functional equivalents, a substantial number are rendered literally. This leads to some problems in accuracy culturally and meaningfully. The study underscores the need for enhancing NMT systems with idiom-specific modules and culturally enriched datasets to improve translation quality.