Wisdom Across Time: NLP-Assisted Translation of Al-Mutanabbi’s Aphoristic Poetry from Arabic into English

Abstract

Objectives: This study investigates the application of natural language processing (NLP) to the translation of aphoristic wisdom poetry by the renowned poet Al-Mutanabbi from Classical Arabic into idiomatic English. It compares NLP-augmented translations with human translations in terms of semantic accuracy, faithfulness, poetic quality, metaphor rendering, cultural reference, and prosody.

Methods: Five poetic lines emphasizing wisdom were selected. The texts were preprocessed through tokenization, lemmatization, and morphological analysis. Initial translations were generated using a transformer–decoder model and subsequently refined by a bilingual literary scholar. Evaluation combined quantitative metrics (BLEU, METEOR, semantic similarity) with qualitative expert judgments addressing style, metaphor realization, and cultural sensitivity.

Results: Raw system output showed moderate performance (BLEU = 0.35; METEOR = 0.42). Post-editing significantly improved results (BLEU = 0.68; METEOR = 0.75). Expert evaluations confirmed notable improvements in poetic aesthetics (2.1 → 3.9), metaphor interpretation (1.8 → 3.5), and cultural nuance (2.0 → 3.7). Analysis indicated that human intervention was required in 70% of metaphor cases and 85% of cultural references.

Conclusions: The findings indicate that while NLP can capture structural patterns and basic meaning, substantial challenges remain in conveying deeper poetic and cultural layers. An integrated approach combining computational efficiency with human expertise yields translations that balance accuracy with literary and cultural authenticity, offering a promising model for translating classical poetry.