Gender Bias in English-to-Ukrainian Machine Translation

This study investigates gender bias in English-to-Ukrainian machine translation by analyzing translations generated by Google Translate and DeepL. The analysis examines how these systems handle genderneutral source texts, particularly in professional, pronoun-related, and stereotypically gendered contexts. The methodology is grounded in a corpus-based mixed-methods approach, which combines a qualitative analysis of a manually compiled mini-corpus with a quantitative comparison against frequency data from the General Regionally Annotated Corpus of Ukrainian (GRAK). It was found that both systems frequently default to masculine forms, especially for high-status and technical professions. This bias reinforces social and cultural stereotypes, and the resulting translations often diverge from natural gender distributions in the Ukrainian language. The fi ndings suggest that without specifi c interventions, machine translation can perpetuate gender stereotypes. Key words: machine translation gender bias, English-Ukrainian translation, corpus, Google Translate, DeepL, gendered language

Gender bias · Machine translation · Qualitative analysis · Qualitative research · Ukrainian · Authorship Attribution and Profiling · Gender Studies in Language · Linguistics, Language Diversity, and Identity