Search papers, labs, and topics across Lattice.
This paper investigates gender recoverability in English-to-Hindi translation, identifying a significant issue where translation systems erase gender cues through ergative and honorific constructions. To address this, they introduce two inference-time interventions: a Source-Aware Reranker (SAR) to avoid gender-neutralizing syntax and a Phenomenon-Aware Reranker (PAR) to preserve gender through targeted lexical marking. Experiments on a 37K instance benchmark show that PAR substantially improves gender preservation, but at the cost of reduced fluency, highlighting a trade-off between cultural fidelity and stylistic naturalness.
English-to-Hindi translation often neuters gender cues, but a simple reranking trick can boost gender preservation by 70%鈥攖hough fluency takes a hit.
Generative translation systems are cultural technologies because they decide how socially meaningful cues are rendered within culturally specific grammatical systems. We study one concrete notion of successful cultural translation: when an English source explicitly encodes gender, an English-to-Hindi translation should preserve the recoverability of that cue unless the source itself is ambiguous. We evaluate this criterion on a 37,345-instance benchmark spanning twelve categories and show that five systems frequently erase gender through ergative and honorific constructions. We then introduce two mechanism-aware inference-time interventions. The first, the Source-Aware Reranker (SAR), prefers candidates that avoid gender-neutralizing syntax. The second, the Phenomenon-Aware Reranker (PAR), preserves gender through targeted lexical marking even when ergative syntax remains. Across GPT-4o-mini and Sarvam, PAR improves target-subset accuracy from 11.07% to 54.47% and from 15.99% to 49.66%, respectively. Human evaluation shows that PAR increases gender preservation from 10.3% to 81.3%, but reduces mean fluency from 4.36 to 3.37. These findings place the two interventions on a preservation and fluency frontier rather than supporting a single dominant solution, and show how culturally situated generation can require explicit tradeoffs among fidelity, fluency, and stylistic naturalness.
