Stylistic analysis of translated languages: A perturbation-based XAI deep learning framework

Dan Feng Huang, Dennis Tay · 2026

Read Stylistic analysis of translated languages: A perturbation-based XAI deep learning framework by Dan Feng Huang, Dennis Tay (2026) 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/plos-pone-0352889.txt.

Description

Dan Feng Huang and Dennis Tay present this 2026 PLOS One study on the distinctive style of translated language, using speeches from United Nations meetings as their test case. Translated text carries subtle fingerprints that differ from originally written text, and earlier classifiers could spot them without explaining how. The authors built a deep learning framework designed to reveal its own reasoning through a perturbation method that tests which features drive its decisions.

Their model separated translated from non-translated texts with strong accuracy. More importantly, the explanation step showed the decisions rested on style patterns deeper than surface word choice or sentence structure. The authors offer the framework as a proof of concept for other language studies where researchers need to know not just what a model decides, but why.

Catalog record

CREATORDan Feng Huang, Dennis Tay
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE46868
LICENSECC_BY
RIGHTSCC BY
SOURCEPLOS

Source: https://doi.org/10.1371/journal.pone.0352889

Browse the full library · About the Tower of Babel