Translation Style Variation across Registers in Machine Translation and Large Language Model Outputs
A Corpus-Based Study of English-to-Chinese Translation
Abstract
This study examines translation style variation across registers in English Chinese outputs from TencentMT, DeepSeek, and ArkLLM. Using an aligned corpus of public speech, legal-diplomatic texts, and news, it compares length ratio, mean sentence length, and connective density through mixed-effects modelling and qualitative analysis. ArkLLM generally produced longer outputs, while TencentMT produced shorter sentences and used more explicit connectives. Register also shaped these patterns, showing that AI translation style varies across systems and communicative contexts.
Submitted: April 30, 2026 | Accepted: Aug. 3, 2026 | Published Oct. 12, 2026 | Language: en
Keywords English-to-Chinese translation • Large language models • Translation style • Register variation • Machine translation
Copyright © 2026 Shuyue Wang, Gang Cheng. This is an open-access work distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction is permitted, provided that the original author(s) and the copyright owner(s) are credited and that the original publication is cited, in accordance with accepted academic practice. The license allows for commercial use. No use, distribution or reproduction is permitted which does not comply with these terms.