Cultural Fidelity in English-to-Hindi Translation: A Preservation-Fluency Frontier for Gender Recoverability

2026-05-26

AI Summary

This study addresses the loss of explicit gender information in English-to-Hindi machine translation due to divergent grammatical structures—particularly ergativity and honorifics—which undermines cultural fidelity. For the first time, gender recoverability is introduced as a core metric for evaluating cultural faithfulness. The authors propose two inference-time, phenomenon-aware reranking strategies: a Source-Aware Reranker (SAR) and a Phenomenon-Aware Reranker (PAR), both leveraging large language models (GPT-4o-mini and Sarvam) to intervene in the translation process. Experimental results demonstrate that PAR substantially improves gender accuracy on targeted subsets from 11%–16% to 50%–55%, while human evaluations show a dramatic increase in gender retention from 10.3% to 81.3%, highlighting a critical trade-off between translational fluency and fidelity to source-gender semantics.

Abstract

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.

Problem

• cultural fidelity
• gender recoverability
• machine translation
• English-to-Hindi
• gender erasure

Innovation

• cultural fidelity
• gender recoverability
• mechanism-aware reranking
• ergative syntax
• translation intervention