Mapping Self-Translation’s Stylistic Distinctiveness: A Comparative Analysis of Human and LLM Translations of Chun-Chan Yeh’s The Mountain Village

Self-translator merges the roles of author and translator, offering unique insights into creative and linguistic choices. As large language models (LLMs) increasingly perform translation tasks, questions arise about how machine-generated translations compare stylistically to human-authored ones. This study examines the stylistic features of self-translation by comparing Chun-Chan Yeh’s Chinese self-translation of his English novel with an allograph translation and two LLM-generated translations (ChatGPT-4o and DeepSeek V3, prompted to simulate a self-translator persona). Using corpus-based analysis of syntactic complexity and phraseological patterns, complemented by qualitative textual analysis, this study finds that Yeh’s self-translation displays greater structural complexity, more culturally embedded phraseological choices, and stronger stylistic control than either the allograph or LLM-generated translations. The comparison with the allograph version further highlights self-translation’s distinctive qualities of stylistic flexibility and strategic depth. This research enhances our understanding of self-translators’ style and the role and functions of AI in literary translation.