The advancement of Artificial Intelligence (AI) has brought innovations and concerns to the translation field. Although previously more commonly used Neural Machine Translation (NMT) tools offer substantial help to translators, Large Language Models (LLMs) introduce new approaches to translation by leveraging vast knowledge bases and focusing on wider context (sentences, paragraphs) rather than word-for-word translation.
However, although efficient in technical texts, machine translation tools face challenges when it comes to political or cultural translation. In these cases, the intervention of the human agent is considered crucial for the accuracy of the outcome so that the translated text would bear the cultural, political, and religious factors of the original. The present research aims to examine two questions: a. Which translation tool provides better assistance in the Turkish-Greek language pair regarding political and religious texts? b. To what extent do specific instruction prompts affect the accuracy of the translation and the religious and cultural terms and factors included in the text? Is there intervention regarding the promotion of certain ideologies or established norms?
To examine the first research question, extracts from the Turkish book “Bir Fikir ve Aksiyon İnsani Olarak - Fethullah Gülen” dealing with issues of religion, politics, and culture were inserted into an NMT tool (Google Translate), a hybrid tool (DeepL Pro NextGen) using both NMT and LLM technologies, and an LLM tool (Chat GPT Pro). The results were examined using the MQM error typology and, more specifically, according to the terminology, accuracy, linguistic conventions, style, and audience appropriateness parameters, with a qualitative Expert Bilingual Review. For the second question, selected extracts were inserted into the LLM tool with different instruction prompts (e.g., translate for readers who have no knowledge of Islam, who are Muslims, who are secular or who are Islamophobic), and the outcomes were analyzed with regard to translation style and additions/omissions. For this stage, the Critical Discourse Analysis framework was also used as a tool for the detection of ideologies and established ideas reproduced by the system.
Iralia Kotsakidou is an adjunct lecturer in the Faculty of Turkish and Modern Asian Studies of the National and Kapodistrian University of Athens in the field of text and discourse analysis and political, economical and social translation (Turkish, English, Greek) and a post-doctoral researcher in the Department of Foreign Languages, Translation and Intrerpreting of Ionian University, where she conducts research regarding the Hizmet movement religious community in Greece. She holds a PhD in Political Discourse Analysis in Turkey from the same department, which was funded by the 3rd Call for HFRI PhD Fellowships. She has attended scientific conferences and written academic articles regarding Critical Discourse Analysis in Recep Tayyip Erdoğan’s speeches, the use of emotionalism in politics, the Hizmet movement and its action in Turkey and worldwide and the challenges of translating cultural, religious and political texts. She has also translated a book from Turkish to Greek called “Φετχουλλαχ Γκιουλεν: Ένας άνθρωπος ιδεών και δράσης» and works as an official translator.
Minas Pergantis is an adjunct lecturer and postdoctoral researcher in the field of User Experience (UX) and Interface Design at the Department of Audio & Visual Arts of the Ionian University where he is conducting research focusing on UX evaluation and optimization of online digital art repositories. He is active in the field of applied Artificial Intelligence including expertise on integrating algorithmic translation in the user and content management experience. He holds a PhD from the same department for his research in the fields of online Art search, user behavior analysis and search technologies. He also holds an integrated Master's in Electrical and Computer Engineering from the University of Patras. As an academic researcher, he specializes in the fields of Human-Computer Interaction (HCI) and usability, applied Artificial Intelligence, Web and Semantic Web data extraction and computational analysis, search engines and SEO, and user behavior analysis based on qualitative and quantitative data, always with an emphasis on Art and Cultural Heritage applications. In addition to his research activity, he is a professional software engineer with many years of experience in the private sector in the fields of website, Web application and smartphone application design and development.
