AI-Driven and LLM-Based Translation of Arabic News Texts into English: A Comparative Evaluation

Abstract

The advancement of artificial intelligence (AI) has significantly impacted the field of translation, introducing new possibilities through AI-driven translation tools and large language models (LLMs). The study evaluated the performance of selected AI-driven translation tools—Google Translate, Reverso, and Yandex—and LLMs—Chat GPT-4, Gemini-1.5-Pro, and Bing—in translating Arabic news texts into English. A quantitative research design was employed, using a corpus of twenty Arabic news texts sourced from Al-Jazeera Net, Russia Today, Al-Quds Al-Arabi, Asharq Al-Awsat, Marebpress, Alarabia Net, BBC Arabic. Translation errors were categorized into lexico-semantic, syntactic, and formatting types. Lexico-semantic errors were the most frequent (45.22%), followed by formatting and syntactic errors (32.27%) (22.50 %) respectively. Among all tools, Chat GPT-4 exhibited the lowest number of errors across all categories (19 out of 471), while Reverso recorded the highest (128 out of 471). Performance scores confirmed that Chat GPT-4 significantly outperformed the other tools, with Reverso scoring the lowest. Regarding translation accuracy, Chat GPT-4 ranked highest, followed by Bing. The remaining tools struggled to produce highly accurate translations of the selected news texts. These findings highlight the increasing effectiveness of (LLMs) in translating domain-specific content and emphasize the need for human post-editing to ensure the accuracy in professional translation settings.

Ähnliche Arbeiten