Cohesion-based Errors and Intelligibility of Persian-English Machine Translation Output

Abstract

Basically, nowadays, such translation-based computer technologies as machine translation systems are employed so as to increase the speed and achieve far higher quality in translation due to the high volume of translation projects and limited time periods. The main goal of this research is to identify the cohesion- based errors committed in machine translation, to evaluate their impact upon the chains of the cohesive devices and also upon the intelligibility of machine translated outputs. In this research, the data was collected from three different machine translation systems, followed by 30 master students of TEFL as the participants in this research were employed in order to back translate a selected quantity of machine translated texts so as to evaluate the quality and the level of intelligibility. Based on the obtained results, it was determined that Google machine translation system assigned the first position due to its better quality and the less proportion of cohesion-based errors committed. On the other hand, Abadis, with a relatively small distinction compared to Google's performance in terms of the quality and the statistic of committed errors, allocated the second position, which confirmed its far better performance than Bing. In addition, regarding the cohesion-based errors affecting the intelligibility of the machine translated texts, it could be argued that among the six types of cohesion-based errors identified, the missing-word error and the non-translated word error were taken into account with the highest and the lowest impact upon the participants' intelligibility of the machine translated texts.