Artificial Intelligence based Arabic-to-English machine versus human translation of poetry: An analytical study of outcomes

Abstract

This study discusses the discrepancies between a machine translation and a human translation of Arabic poems to English. The dataset comprises two Arabic poems, Nothing of Note, and One Day. Each of these have English translations achieved by one or more translators working in unison to ensure close translation of the original. The researcher then generated the Artificial Intelligence (AI)-based Machine Translation (MT) versions of these and conducted a critical linguistic examination of the two outputs. Results indicated that the MT fails to capture the cultural context for instance, in the poem, Nothing of Note, the protagonist is uprooted his salary is changed to "pounds," and the use of different adjectives in the MT creates a different meaning than the original. The human translation's use of conjunctive pairs in the same creates a lyrical continuance to the adjectival antonyms, which is not achieved in the MT. The poem's context is also lost in the MT, e.g., "the dust-dunned street" is changed to "dusty street," and "like time would not walk with him" is changed to "on the ground, as if he was walking, but time wasn't passing." Finally, MT establishes two protagonists, while the human translation does not. In the second analysis of the poem, One Day, the limitations of machine translation are stark in capturing the socio-cultural context of poetry. The critical linguistic analysis comparing human translation to MT, points out that the latter failed to capture the nuances of the poem, including the use of figurative language, historical references, and the genre of the poem. In conclusion, MT is unable to apply meaning beyond its database and lacks the ability to understand the cultural context in which the poetry was created and can therefore, not be a good tool for translation of Arabic poetry to English.