The English Trap: Gender Bias and Grammatical Information Loss Through English-Influenced Universal Representations in Multilingual NMT
Abstract
Dataset and code accompanying the paper "The English Trap" (2026). Contains 112 annotated sentences across 3 MT systems and 2 translation directions (Spanish→Greek, Greek→Spanish), with gender bias annotations, justifications, error type classifications, full pivot language translations, back-translations, and exact-match scores for 16 candidate pivot languages. Version 2.0: complete dataset replacing v1 which had incomplete pivot scores.
Fields of Study & Subjects
Natural Language Processing Techniques
Translation Studies and Practices
Topic Modeling
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 18 February 2026
- Language
- EN
- Document type
- Dataset