Challenges and strategies in post-editing English into Arabic Neural Machine Translations of movie subtitles

Abstract

This study investigates the challenges and strategies involved in post-editing (PE) Neural Machine Translation (NMT) subtitles for the Netflix movie La La Land from English to Arabic, utilizing Gottlieb's ten subtitling strategies. By adopting a descriptive-qualitative approach, this study identified issues related to linguistic fidelity, cultural adaptation, and technical constraints. The findings also revealed an enhanced understanding of audiovisual translation workflows, emphasizing the role of human expertise in refining machine-generated outputs. Among these, paraphrasing was the most frequently used strategy as many dialogues required localization and modification to enhance clarity for an Arabic audience. In contrast, transcription, cultural substitution, and deletion were employed less often. The study highlights the importance of human expertise in refining machine-generated output and offers recommendations for optimizing Arabic subtitle PE in NMT environments.