Human vs. AI: A Comparative Analysis of Translating Near-Synonymy in the Quran

Abstract

Near-synonymy in the Qur’an challenges translators because related lexical items can differ in intensity, scope, expressive force, and contextual function. This exploratory qualitative study compares selected English expressions by Arberry and Irving with excerpts reported as GPT-4 outputs. Eleven lexical items in ten verse contexts are examined through a translation-oriented adaptation of Murphy’s Relation by Contrast approach. The analysis draws on published, tafsir-informed semantic profiles and distinguishes local semantic adequacy, preservation of lexical contrast, contextual fit, and English idiomaticity. The results vary by item. All three products maintain lexical differentiation between bathth and ḥuzn in Qur’an 12:86. Irving’s ‘famine’ and the AI expression ‘severe hunger’ express different aspects of the masghabah/jūʿ contrast. Arberry’s ‘grievous sin’ makes religious condemnation explicit, while ‘fault’ remains defensible for khaṭīʾah. The AI excerpts make regret and private speech explicit, but ‘despair impatiently’ is an unusual collocation. These findings support evaluation by specific semantic features rather than an overall human-AI ranking. The contribution is an account of how lexical contrast, interpretive choice, and English usage interact in selected translation examples. Incomplete AI provenance, excerpt-level evidence, and the absence of independent ratings limit the conclusions. A supervised revision workflow is proposed for subsequent testing.