Sentence splitting in Arabic to English and Spanish translation: a statistical based, training-oriented study

Abstract

In Modern Standard Arabic punctuation marks are less frequent than in European languages. Using discourse markers is instead preferred, which tends to yield long clauses. This, in turn, is assumed to foster sentence splitting in translation from Arabic into English and into Spanish. This assumption was tested on a sample of 470 554 words from a trilingual corpus of 14 569 764 words. Our results show that each Arabic sentence yielded an average of 1.5 English and Spanish sentences. Our data also reflects the impact of directionality, because sentence splitting is almost non-existent in translation from English and from Spanish into Arabic. We also statistically determined the most salient markers for sentence splitting. These findings are particularly relevant to translator training. We provide examples of real-world translations showing levels of sentence splitting above or below average and suggest training activities aimed at raising awareness of this aspect of translation among, but not exclusively, English or Spanish L1/Arabic L2 trainees. This material can also be used for future research.