From Binary Defaults to Contextual Bias: Translating Queer Morphology with NMT and LLMs

A study by Manuel Lardelli, published in the Proceedings of the 4th Workshop on Gender-Inclusive Translation Technologies (GITT 2026), found that neural machine translation (NMT) systems default to binary masculine/feminine grammar and flip inconsistently between forms for the same subject, erasing queer visibility, while large language models (LLMs) attempt gender-fair language through neutralization and neomorphemes such as the schwa but introduce systematic errors including contextual over-feminization and structurally invalid word endings. The analysis covered 12 NMT and 15 LLM translations of German literary fiction into Italian from a human-in-the-loop experiment. The findings, published in Tilburg, the Netherlands, pages 31–48 by the European Association for Machine Translation, offer preliminary empirical guidance for post-editors handling gender-fair translation.

Abstract This paper evaluates how Neural Machine Translation NMT and Large Language Models LLMs process non-binary morphology when translating German literary fiction into Italian. We apply an inductive, mixed-methods framework to analyze 12 NMT and 15 LLM translations from a human-in-the-loop experiment. Results reveal a fundamental divergence. NMT defaults to standard binary grammar but applies it inconsistently, often flipping between masculine and feminine forms for the same subject across different sentences, which effectively erases queer visibility. Conversely, LLMs actively attempt gender-fair language via neutralization and neomorphemes e.g., the schwa . However, LLMs introduce new systematic errors: driven by semantic cues, they exhibit a contextual bias that, in the present study, frequently led to over-feminization, and their attempts to create inclusive word endings result in structurally invalid words. Ultimately, these findings expose current limitations and provide preliminary empirical guidance to assist post-editors in navigating the complex challenges of gender-fair translation. - Anthology ID: - 2026.gitt-1.4 - Volume: - Proceedings of the 4th Workshop on Gender-Inclusive Translation Technologies GITT 2026 https://aclanthology.org/volumes/2026.gitt-1/ - Month: - June - Year: - 2026 - Address: - Tilburg, the Netherlands - Editors: - Manuel Lardelli https://aclanthology.org/people/manuel-lardelli/ , Beatrice Savoldi https://aclanthology.org/people/beatrice-savoldi/ , Janiça Hackenbuchner https://aclanthology.org/people/janica-hackenbuchner/ , Luisa Bentivogli https://aclanthology.org/people/luisa-bentivogli/ , Eleni Gkovedarou https://aclanthology.org/people/eleni-gkovedarou/ , Joke Daems https://aclanthology.org/people/joke-daems/ - Venues: - GITT https://aclanthology.org/venues/gitt/ | WS https://aclanthology.org/venues/ws/ - SIG: - Publisher: - European Association for Machine Translation - Note: - Pages: - 31–48 - Language: - URL: - https://aclanthology.org/2026.gitt-1.4/ https://aclanthology.org/2026.gitt-1.4/ - DOI: - Cite ACL : - Manuel Lardelli. 2026. From Binary Defaults to Contextual Bias: Translating Queer Morphology with NMT and LLMs https://aclanthology.org/2026.gitt-1.4/ . In Proceedings of the 4th Workshop on Gender-Inclusive Translation Technologies GITT 2026 , pages 31–48, Tilburg, the Netherlands. European Association for Machine Translation. - Cite Informal : - From Binary Defaults to Contextual Bias: Translating Queer Morphology with NMT and LLMs https://aclanthology.org/2026.gitt-1.4/ Lardelli, GITT 2026 - PDF: - https://aclanthology.org/2026.gitt-1.4.pdf https://aclanthology.org/2026.gitt-1.4.pdf