Document Type : Research Paper
Author
Assistant Professor of Translation Studies, English Language Department, Hazrat-e Masoumeh University, Qom, Iran
Abstract
Given that adherence to the orthographic principles and conventions of the target language constitutes a significant dimension of translation quality, this study investigated the compliance of English-to-Persian journalistic translations generated by five online machine translation (MT) systems with guidelines of the approved orthography of the Academy of Persian Language and Literature (APLL) and standard usage rules for punctuation marks. To this end, several texts were extracted from English newspapers and were translated into Persian by five systems (ChatGPT, DeepSeek, Google Translate, Microsoft Translator, and Reverso). Through meticulous, word-by-word reading and analysis of the translations in the light of the guidelines of the Approved orthography and usage rules for punctuation marks, a total of 16 distinct categories of orthographic and punctuation errors were identified in the machine-generated translations. These were later classified into three main types: (1) errors in the use of non-breaking or half-space (Nim-Faslé) within word elements and elements of compound words, (2) errors in the orthography of secondary Persian characters, and (3) errors in the correct spacing of punctuation marks. Google, Microsoft, and Reverso produced instances of all error types. DeepSeek and ChatGPT exhibited 4 and 12 types of these errors in their outputs, respectively. Furthermore, the results indicated that none of the systems demonstrated complete consistency in performance so that a given compound or character was at times rendered correctly and other times incorrectly. It was, also, discerned that the identified errors bear a significant resemblance to orthographic errors prevalent in human-generated texts. This phenomenon is attributable to the predominantly open-source monolingual and bilingual corpora used for training these online systems, which comprise texts authored and translated by diverse individuals across varied contexts and registers. Considering the increasing reliance on online MT systems, such orthographic and punctuation errors possess the potential to gradually permeate into the texts generated by Persian-speaking communities, potentially undermining the APLL's ongoing orthographic standardization efforts. One proposed solution for this problem entails the development, either by the MT providers themselves or independent developers, of dedicated or integrated auxiliary systems designed to scrutinize and post-edit translations prior to final user delivery, specifically evaluating their conformity with the approved orthographic guidelines and standard punctuation conventions.
Keywords
Main Subjects
)