Research on the Translation of Chinese Complex Sentences With Large Language Model ChatGPT-4o —Taking English Translation of Chinese Flowing Sentences as an Example

Large language models (LLMs) have demonstrated outstanding performance in the field of machine translation. Many scholars have explored LLM-based translation from various perspectives, primarily focusing on the translation of routine, simple, and everyday texts. However, when it comes to the examination of specific sentence structures, especially complex sentences, the research is still somewhat insufficient. Therefore, this study attempts to investigate the translation of Chinese complex sentences by ChatGPT-4o, focusing on the translation of Chinese flowing sentences, a special sentence pattern in Chinese. The study examines how ChatGPT-4o handles the subject identification and translation of zero-subject sentence segments within Chinese flowing sentences, as well as the translation of the frozen sentence segments in Chinese flowing sentences. The research results indicate that: for the ten categories of subject identification in zero-subject sentence segments, ChatGPT-4o can generally identify them all, but occasionally makes ambiguous identification; ChatGPT-4o employs two translation strategies—one is to add subjects (mostly in the form of pronouns) to translate the zero-subject sentence segments into independent clauses, the other is to translate the zero-subject sentence segments into attributive structures; the accuracy of ChatGPT-4o’s translation of frozen sentence segments is inconsistent; ChatGPT-4o may adopt the literal meaning or metaphorical meaning of the frozen sentence segment, determining the form of the frozen sentence segment based on the overall structure of the flowing sentence.