A Linguistic Challenge to ChatGPT: Use of AI in Detection, Text Generation and Translation of False Friends in Turkish-English

Authors : Evren Barut

Pages : 154-177

Doi:10.29110/soylemdergi.1598386

View : 89 | Download : 49

Publication Date : 2025-03-25

Article Type : Research Paper

Abstract :This study aims to analyze AI-supported text generation and translation technology, focusing on the linguistic challenge of false friends. False friends can cause semantic difficulties in both language learning and translation. The study examined the capabilities of AI tools in handling false friends as a linguistic challenge, using GPT-4o as the AI model. The analysis consisted of two phases. In the first phase, GPT-4o processed false friends in the English-Turkish language pair and generated a text without relying on any predefined dataset. In the second phase, Nihal Yetkin’s (2011) study on 100 false friends in Turkish and English within the field of diplomacy was used as a dataset. Using this thematic data, GPT-4o was tasked with generating a text with a limited dataset. Both AI-generated texts were then translated using GPT-4o. The study evaluated GPT-4o’s linguistic creativity and data processing skills in addressing the complexities of false friends. The findings discussed how GPT-4o recognized false friends and used them contextually in text generation. The primary objective in both phases was to assess GPT-4o’s ability to use false friends in appropriate contexts and its performance in translating these words into the target language. The analysis of findings provided insights into the effectiveness of AI-powered translation in handling linguistic challenges. Ultimately, this study assessed GPT-4o’s success in identifying and processing Turkish-English false friends in text generation and translation.
Keywords : Çeviribilim, Yalancı eşdeğer, Anlambilim, Dilbilim, OpenAI, GPT-4o, Yapay Zeka Metin Üretimi, Yapay Zeka Destekli Çeviri