An entropy-based study of Simplification in ChatGPT translations compared to neural machine translation and human translation across genres
Guangyuan Yao, Lingxi Fan · 2025
Read An entropy-based study of Simplification in ChatGPT translations compared to neural machine translation and human translation across genres by Guangyuan Yao, Lingxi Fan (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-12755776.txt.
Description
Machine translation often flattens rich language into simpler wording, and a study asked whether a popular AI chatbot flattens Chinese-to-English translation even more. Guangyuan Yao and Lingxi Fan measured that flattening in this 2025 study using entropy scores. They compared human work, a standard neural system, and the chatbot across three text types.
Word-level and grammar-level entropy tracked how varied the wording and sentence shapes stayed. Political and academic texts lost word variety in machine versions, and all text types lost grammar variety. The degree of loss differed by system, with genre playing a smaller role for word choice. The results matter for readers who need faithful tone, since quiet simplification can drain legal or literary texts of meaning.
Catalog record
| CREATOR | Guangyuan Yao, Lingxi Fan |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |