Missing the human touch? A computational stylometry analysis of GPT-4 translations of online Chinese literature
Xiaofang Yao, Yong-Bin Kang, Anthony McCosker
TL;DR
The paper investigates whether GPT-4 can replicate the stylistic fingerprint of human translations in Chinese online literature by applying computational stylometry to a large open parallel corpus. Using two context-aware prompting strategies and a dataset of 25 novels from the BWB corpus, it compares GPT-4 outputs against human translations along lexical, syntactic, and content features. Results show GPT-4 largely aligns with human translations across many stylistic dimensions, though human translators retain greater lexical diversity and challenges remain in semantic interpretation, proper-noun transliteration, and polysemy. The study advances posthumanist translation discourse through reusable prompts and an open parallel corpus, and highlights how AI can augment, rather than replace, human translators in literary contexts while outlining areas needing careful oversight.
Abstract
Existing research indicates that machine translations (MTs) of literary texts are often unsatisfactory. MTs are typically evaluated using automated metrics and subjective human ratings, with limited focus on stylistic features. Evidence is also limited on whether state-of-the-art large language models (LLMs) will reshape literary translation. This study examines the stylistic features of LLM translations, comparing GPT-4's performance to human translations in a Chinese online literature task. Computational stylometry analysis shows that GPT-4 translations closely align with human translations in lexical, syntactic, and content features, suggesting that LLMs might replicate the 'human touch' in literary translation style. These findings offer insights into AI's impact on literary translation from a posthuman perspective, where distinctions between machine and human translations become increasingly blurry.