Adaptive Translation of English to Arabic Movie Subtitles Using GPT-3.5 Turbo and FAISS
Wajiha Abdul Shakir · 2024
This research explores adaptive machine translation, focusing on the translation of English-to-Arabic movie subtitles using advanced large language models such as GPT-3.5 Turbo. Utilizing Sentence-Transformer to generate embeddings and FAISS for efficient retrieval enhances the translation quality, particularly in preserving idiomatic expressions and contextual nuances. The evaluation metrics—BLEU, CHRF, and TER—show significant improvements, with BLEU increasing by up to 12% over baseline systems. These results underscore the potential of adaptive techniques for enhancing subtitle translations in real-time, especially for complex languages like Arabic. The findings point towards the future development of more effective, domain-specific translation applications.