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

Chavi Gupta; Dhruv Kumar; Samyak Savi; Shreyas Gantayet; Tanay Sodha

Preserving explicit gender in English-to-Hindi translation requires trading fluency for recoverability via targeted rerankers.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.3

2026-06-29 18:09 UTC pith:VST64EVY

load-bearing objection PAR lifts gender recoverability in English-to-Hindi MT on a new benchmark but trades off fluency, with the benchmark's construction as the main open question. the 2 major comments →

arxiv 2605.27654 v1 pith:VST64EVY submitted 2026-05-26 cs.CL cs.AIcs.CY

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

Samyak Savi , Chavi Gupta , Shreyas Gantayet , Tanay Sodha , Dhruv Kumar This is my paper

classification cs.CL cs.AIcs.CY

keywords gender recoverabilityEnglish-to-Hindi translationcultural fidelitymachine translationinference-time rerankingfluency tradeoffergative constructionshonorifics

verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper sets out to show that when an English source makes gender explicit, a translation into Hindi should keep that information recoverable rather than letting it disappear through neutral syntax. Five current systems often erase the cue via ergative and honorific constructions on a benchmark of over 37,000 examples. The authors introduce two inference-time rerankers: Source-Aware Reranker avoids neutralizing syntax, while Phenomenon-Aware Reranker forces lexical gender marking. Tests on GPT-4o-mini and Sarvam show the second method raises gender accuracy on the relevant subset from roughly 11-16 percent to 50-54 percent, with human raters confirming preservation rising from 10 percent to 81 percent, though at a clear drop in fluency. The result frames the task as selecting a point on a preservation-fluency frontier instead of claiming one superior approach.

Core claim

When English explicitly encodes gender, an English-to-Hindi translation should preserve the recoverability of that cue unless the source itself is ambiguous. Five systems frequently erase gender through ergative and honorific constructions on a 37,345-instance benchmark spanning twelve categories. The Phenomenon-Aware Reranker preserves gender through targeted lexical marking even when ergative syntax remains, improving target-subset accuracy from 11.07 percent to 54.47 percent for GPT-4o-mini and from 15.99 percent to 49.66 percent for Sarvam, while human evaluation shows gender preservation rising from 10.3 percent to 81.3 percent at the cost of mean fluency falling from 4.36 to 3.37.

What carries the argument

The Phenomenon-Aware Reranker (PAR), an inference-time intervention that preserves gender through targeted lexical marking even when ergative syntax remains.

Load-bearing premise

That preserving explicit gender recoverability is the primary criterion for successful cultural translation.

What would settle it

A controlled rating study in which native Hindi speakers consistently prefer higher-fluency gender-neutral translations over lower-fluency gender-preserving ones on the same inputs would undermine the priority assigned to recoverability.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

PAR raises target-subset accuracy on gender-encoding inputs for both GPT-4o-mini and Sarvam.
Human evaluations record large increases in gender preservation rates.
Mean fluency scores decline as a direct consequence of the intervention.
The two rerankers place outputs on a preservation-fluency frontier with no single dominant solution.
Cultural translation can require explicit tradeoffs among fidelity, fluency, and stylistic naturalness.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same reranking logic could be applied to other source-language features such as number or social register in languages with complex morphology.
The twelve-category benchmark offers a diagnostic tool for identifying which English constructions most reliably trigger erasure.
Translation systems could expose user controls that let speakers choose priority between recoverability and fluency on demand.
Testing the rerankers on additional language pairs that mark gender or honorifics would clarify how far the tradeoff pattern generalizes.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Desk Editor's Note

PAR lifts gender recoverability in English-to-Hindi MT on a new benchmark but trades off fluency, with the benchmark's construction as the main open question.

read the letter

The main takeaway is that this paper introduces a 37k-instance benchmark for gender recoverability in English-to-Hindi translation and shows that their PAR reranker raises target-subset accuracy from the low teens to around 50% on two systems while boosting human-rated preservation from 10% to 81%, at the cost of lower fluency.

What stands out as new is the specific application of mechanism-aware reranking to ergative and honorific neutralization. SAR avoids certain syntax; PAR adds lexical markers to keep gender cues recoverable. They evaluate five systems, report the quantitative lifts, and include a human study that documents the preservation-fluency frontier. That framing avoids overclaiming and treats the output as a cultural decision.

The soft spot is the benchmark. The twelve categories and 37k size are concrete, but the abstract and stress-test note leave the selection of English sources unclear. If the inputs over-represent certain pronouns, names, or occupational cases, the reported gains may not generalize to typical gender-encoding text. Full methods on how the benchmark was built and any category breakdown would clarify this.

The work is aimed at MT researchers handling grammatical gender and cultural fidelity in Indic languages. It has enough empirical grounding and a clear criterion to deserve peer review, though a referee would likely press on benchmark validation and additional baselines.

I would send it out rather than desk reject.

Referee Report

2 major / 0 minor

Summary. The paper claims that English-to-Hindi MT systems frequently erase explicit gender cues from sources via ergative and honorific constructions. It introduces two inference-time rerankers (SAR and PAR) and evaluates them on a 37,345-instance benchmark across twelve categories, reporting that PAR raises target-subset gender recoverability accuracy from 11.07% to 54.47% (GPT-4o-mini) and 15.99% to 49.66% (Sarvam). Human ratings show PAR boosts gender preservation (10.3% to 81.3%) at the cost of fluency (4.36 to 3.37), positioning the interventions on a preservation-fluency frontier.

Significance. If the benchmark construction and evaluation protocols are sound, the work provides concrete empirical evidence of gender erasure in culturally situated translation and demonstrates practical mechanism-aware interventions with explicit trade-offs. The large-scale benchmark and human evaluation of both accuracy and fluency are strengths that could inform future work on fidelity in MT for languages with grammatical gender.

major comments (2)

[Abstract] Abstract: The central quantitative claims rest on improvements measured against a 37,345-instance benchmark spanning twelve categories, yet no construction details, definition of 'explicitly encode gender' (via pronouns, names, or context), or category breakdown are supplied. This directly affects whether the reported jumps (e.g., 11.07% to 54.47%) can be generalized beyond the sampled encodings.
[Abstract] Abstract (human evaluation paragraph): The reported human ratings (gender preservation 10.3% to 81.3%, fluency 4.36 to 3.37) are load-bearing for the preservation-fluency frontier claim, but the abstract supplies no information on rater count, selection of evaluation subsets, statistical tests, or inter-rater agreement, preventing verification of the trade-off result.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for highlighting the need for greater transparency in the abstract. We will revise the abstract to include concise details on benchmark construction and human evaluation protocols, as these are described more fully in the manuscript body. This addresses the concerns about assessing generalizability and verifying the trade-off claims.

read point-by-point responses

Referee: [Abstract] Abstract: The central quantitative claims rest on improvements measured against a 37,345-instance benchmark spanning twelve categories, yet no construction details, definition of 'explicitly encode gender' (via pronouns, names, or context), or category breakdown are supplied. This directly affects whether the reported jumps (e.g., 11.07% to 54.47%) can be generalized beyond the sampled encodings.

Authors: We agree the abstract should supply more context on benchmark construction to support the quantitative claims. The full manuscript details the benchmark creation, the operational definition of explicit gender encoding (via pronouns, names, and context), and the twelve-category breakdown with instance counts. We will revise the abstract to briefly summarize these elements, including the category span and key construction criteria, so readers can better evaluate generalizability of the reported accuracy improvements. revision: yes
Referee: [Abstract] Abstract (human evaluation paragraph): The reported human ratings (gender preservation 10.3% to 81.3%, fluency 4.36 to 3.37) are load-bearing for the preservation-fluency frontier claim, but the abstract supplies no information on rater count, selection of evaluation subsets, statistical tests, or inter-rater agreement, preventing verification of the trade-off result.

Authors: We concur that the abstract requires additional information on the human evaluation to substantiate the preservation-fluency frontier. The manuscript body specifies rater count, subset selection criteria, inter-rater agreement, and statistical tests confirming the significance of the fluency trade-off. We will update the abstract with a concise statement covering rater numbers, evaluation protocol summary, and note on statistical verification to allow direct assessment of the reported ratings. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical measurements on external benchmark with no fitted predictions or self-referential derivations.

full rationale

The paper reports accuracy improvements and human evaluations from applying SAR and PAR interventions to GPT-4o-mini and Sarvam on a fixed 37,345-instance benchmark. No equations, parameter fitting, or predictions derived from the same data appear. Claims rest on direct measurement against the benchmark and separate human raters rather than any internal definition or self-citation chain. The interventions are described as mechanism-aware but their reported effects are observed outcomes, not tautological. This matches the default case of a self-contained empirical study.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that gender recoverability is a primary measure of cultural fidelity and that the chosen benchmark categories adequately sample the phenomenon.

axioms (1)

domain assumption 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.
Stated as the definition of successful cultural translation in the abstract.

pith-pipeline@v0.9.1-grok · 5796 in / 1297 out tokens · 36745 ms · 2026-06-29T18:09:42.653302+00:00 · methodology

0 comments

read the original 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.

discussion (0)

Reference graph

Works this paper leans on

24 extracted references · 13 canonical work pages · 1 internal anchor

[1]

Attanasio, F

G. Attanasio, F. M. Plaza del Arco, D. Nozza, and A. Lauscher. A tale of pronouns: Interpretability informs gender bias mitigation for fairer instruction-tuned machine translation. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 3996--4014, Singapore, 2023. Association for Computational Linguistics. doi:10....

work page doi:10.18653/v1/2023.emnlp-main.243 2023
[2]

Currey, M

A. Currey, M. Nadejde, R. R. Pappagari, M. Mayer, S. Lauly, X. Niu, B. Hsu, and G. Dinu. MT-GenEval : A counterfactual and contextual dataset for evaluating gender accuracy in machine translation. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 4287--4299. Association for Computational Linguistics, 2022. do...

work page doi:10.18653/v1/2022.emnlp-main.288 2022
[3]

J. Gala, P. A. Chitale, A. Raghavan, V. Dhore, S. Sureja, S. Doddapaneni, A. Bapna, G. Ramesh, A. Kunchukuttan, P. Kumar, et al. IndicTrans2 : Towards high-quality and accessible machine translation models for all 22 scheduled Indian languages. Transactions on Machine Learning Research, 2024

2024
[4]

R. Hada, S. Husain, V. Gumma, H. Diddee, A. Yadavalli, A. Seth, N. Kulkarni, U. Gadiraju, A. Vashistha, V. Seshadri, and K. Bali. Akal badi ya bias: An exploratory study of gender bias in H indi language technology. In Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency, 2024. doi:10.1145/3630106.3659017

work page doi:10.1145/3630106.3659017 2024
[5]

Kirtane and T

blue N. Kirtane and T. Anand. Mitigating Gender Stereotypes in Hindi and Marathi . In Proceedings of the 4th Workshop on Gender Bias in Natural Language Processing . Association for Computational Linguistics , 2022

2022
[6]

Moryossef, R

blue A. Moryossef, R. Aharoni, and Y. Goldberg. Filling Gender and Number Gaps in Neural Machine Translation with Black-box Context Injection . In Proceedings of the First Workshop on Gender Bias in Natural Language Processing . Association for Computational Linguistics , 2019

2019
[7]

NLLB Team , M. R. Costa-juss \`a , J. Cross, O. C elebi, M. Elbayad, K. Heafield, et al. No language left behind: Scaling human-centered machine translation. Technical report, Meta AI, 2022. arXiv:2207.04672

work page internal anchor Pith review Pith/arXiv arXiv 2022
[8]

GPT-4o mini: Advancing cost-efficient intelligence, 2024

OpenAI . GPT-4o mini: Advancing cost-efficient intelligence, 2024. OpenAI product announcement. Accessed: March 2026

2024
[9]

Ramesh, G

K. Ramesh, G. Gupta, and S. Singh. Evaluating gender bias in H indi- E nglish machine translation. In Proceedings of the 3rd Workshop on Gender Bias in Natural Language Processing, pages 16--23. Association for Computational Linguistics, 2021. doi:10.18653/v1/2021.gebnlp-1.3

work page doi:10.18653/v1/2021.gebnlp-1.3 2021
[10]

Robinson, S

K. Robinson, S. Kudugunta, R. Stella, S. Dev, and J. Bastings. MiTTenS : A dataset for evaluating gender mistranslation. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 4115--4124. Association for Computational Linguistics, 2024

2024
[11]

blue A. Sant, C. Escolano, A. Mash, F. De Luca Fornaciari, and M. Melero. The Power of Prompts: Evaluating and Mitigating Gender Bias in MT with LLMs . In Proceedings of the 5th Workshop on Gender Bias in Natural Language Processing , pages 94--139, Bangkok, Thailand, 2024. Association for Computational Linguistics . doi:10.18653/v1/2024.gebnlp-1.7. URL h...

work page doi:10.18653/v1/2024.gebnlp-1.7 2024
[12]

Sarvam translate

Sarvam AI . Sarvam translate. https://www.sarvam.ai/blogs/sarvam-translate, 2024. Accessed: March 2026

2024
[13]

Saunders, R

D. Saunders, R. Sallis, and B. Byrne. Neural machine translation doesn't translate gender coreference right unless you make it. In Proceedings of the Second Workshop on Gender Bias in Natural Language Processing, pages 35--43. Association for Computational Linguistics, 2020

2020
[14]

Saunders, R

blue D. Saunders, R. Sallis, and B. Byrne. First the Worst: Finding Better Gender Translations During Beam Search . In Findings of the Association for Computational Linguistics: ACL 2022 . Association for Computational Linguistics , 2022

2022
[15]

Savoldi, M

B. Savoldi, M. Gaido, L. Bentivogli, M. Negri, and M. Turchi. Gender bias in machine translation. Transactions of the Association for Computational Linguistics, 9: 0 845--874, 2021. doi:10.1162/tacl_a_00401

work page doi:10.1162/tacl_a_00401 2021
[16]

P. Singh. Gender inflected or bias inflicted: On using grammatical gender cues for bias evaluation in machine translation. In Proceedings of the 13th International Joint Conference on Natural Language Processing and the 3rd Conference of the Asia-Pacific Chapter of the ACL: Student Research Workshop, pages 17--23, 2023. doi:10.18653/v1/2023.ijcnlp-srw.3

work page doi:10.18653/v1/2023.ijcnlp-srw.3 2023
[17]

Singh, M

P. Singh, M. Patidar, and L. Vig. Translating across cultures: LLM s for intralingual cultural adaptation. In Proceedings of the 28th Conference on Computational Natural Language Learning, pages 400--418, Miami, FL, USA, 2024. Association for Computational Linguistics. doi:10.18653/v1/2024.conll-1.30. URL https://aclanthology.org/2024.conll-1.30/

work page doi:10.18653/v1/2024.conll-1.30 2024
[18]

Stafanovi c s, T

A. Stafanovi c s, T. Bergmanis, and M. Pinnis. Mitigating gender bias in machine translation with target gender annotations. In Proceedings of the Fifth Conference on Machine Translation, pages 629--638, Online, 2020. Association for Computational Linguistics. doi:10.18653/v1/2020.wmt-1.73. URL https://aclanthology.org/2020.wmt-1.73/

work page doi:10.18653/v1/2020.wmt-1.73 2020
[19]

Stanovsky, N

G. Stanovsky, N. A. Smith, and L. Zettlemoyer. Evaluating gender bias in machine translation. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 1679--1684. Association for Computational Linguistics, 2019. doi:10.18653/v1/P19-1164

work page doi:10.18653/v1/p19-1164 2019
[20]

Tiedemann and S

J. Tiedemann and S. Thottingal. OPUS-MT : Building open translation services for the world. In Proceedings of the 22nd Annual Conference of the European Association for Machine Translation, pages 479--480, 2020

2020
[21]

Vanmassenhove, C

blue E. Vanmassenhove, C. Hardmeier, and A. Way. Getting Gender Right in Neural Machine Translation . In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing . Association for Computational Linguistics , 2018

2018
[22]

B. Yao, M. Jiang, T. Bobinac, D. Yang, and J. Hu. Benchmarking machine translation with cultural awareness. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 13078--13096, Miami, Florida, USA, 2024. Association for Computational Linguistics. doi:10.18653/v1/2024.findings-emnlp.765. URL https://aclanthology.org/2024.findings-e...

work page doi:10.18653/v1/2024.findings-emnlp.765 2024
[23]

J. Zhao, T. Wang, M. Yatskar, V. Ordonez, and K.-W. Chang. Gender bias in coreference resolution: Evaluation and debiasing methods. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 15--20. Association for Computational Linguistics, 2018. doi:10.18653/v1...

work page doi:10.18653/v1/n18-2003 2018
[24]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

[1] [1]

Attanasio, F

G. Attanasio, F. M. Plaza del Arco, D. Nozza, and A. Lauscher. A tale of pronouns: Interpretability informs gender bias mitigation for fairer instruction-tuned machine translation. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 3996--4014, Singapore, 2023. Association for Computational Linguistics. doi:10....

work page doi:10.18653/v1/2023.emnlp-main.243 2023

[2] [2]

Currey, M

A. Currey, M. Nadejde, R. R. Pappagari, M. Mayer, S. Lauly, X. Niu, B. Hsu, and G. Dinu. MT-GenEval : A counterfactual and contextual dataset for evaluating gender accuracy in machine translation. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 4287--4299. Association for Computational Linguistics, 2022. do...

work page doi:10.18653/v1/2022.emnlp-main.288 2022

[3] [3]

J. Gala, P. A. Chitale, A. Raghavan, V. Dhore, S. Sureja, S. Doddapaneni, A. Bapna, G. Ramesh, A. Kunchukuttan, P. Kumar, et al. IndicTrans2 : Towards high-quality and accessible machine translation models for all 22 scheduled Indian languages. Transactions on Machine Learning Research, 2024

2024

[4] [4]

R. Hada, S. Husain, V. Gumma, H. Diddee, A. Yadavalli, A. Seth, N. Kulkarni, U. Gadiraju, A. Vashistha, V. Seshadri, and K. Bali. Akal badi ya bias: An exploratory study of gender bias in H indi language technology. In Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency, 2024. doi:10.1145/3630106.3659017

work page doi:10.1145/3630106.3659017 2024

[5] [5]

Kirtane and T

blue N. Kirtane and T. Anand. Mitigating Gender Stereotypes in Hindi and Marathi . In Proceedings of the 4th Workshop on Gender Bias in Natural Language Processing . Association for Computational Linguistics , 2022

2022

[6] [6]

Moryossef, R

blue A. Moryossef, R. Aharoni, and Y. Goldberg. Filling Gender and Number Gaps in Neural Machine Translation with Black-box Context Injection . In Proceedings of the First Workshop on Gender Bias in Natural Language Processing . Association for Computational Linguistics , 2019

2019

[7] [7]

NLLB Team , M. R. Costa-juss \`a , J. Cross, O. C elebi, M. Elbayad, K. Heafield, et al. No language left behind: Scaling human-centered machine translation. Technical report, Meta AI, 2022. arXiv:2207.04672

work page internal anchor Pith review Pith/arXiv arXiv 2022

[8] [8]

GPT-4o mini: Advancing cost-efficient intelligence, 2024

OpenAI . GPT-4o mini: Advancing cost-efficient intelligence, 2024. OpenAI product announcement. Accessed: March 2026

2024

[9] [9]

Ramesh, G

K. Ramesh, G. Gupta, and S. Singh. Evaluating gender bias in H indi- E nglish machine translation. In Proceedings of the 3rd Workshop on Gender Bias in Natural Language Processing, pages 16--23. Association for Computational Linguistics, 2021. doi:10.18653/v1/2021.gebnlp-1.3

work page doi:10.18653/v1/2021.gebnlp-1.3 2021

[10] [10]

Robinson, S

K. Robinson, S. Kudugunta, R. Stella, S. Dev, and J. Bastings. MiTTenS : A dataset for evaluating gender mistranslation. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 4115--4124. Association for Computational Linguistics, 2024

2024

[11] [11]

blue A. Sant, C. Escolano, A. Mash, F. De Luca Fornaciari, and M. Melero. The Power of Prompts: Evaluating and Mitigating Gender Bias in MT with LLMs . In Proceedings of the 5th Workshop on Gender Bias in Natural Language Processing , pages 94--139, Bangkok, Thailand, 2024. Association for Computational Linguistics . doi:10.18653/v1/2024.gebnlp-1.7. URL h...

work page doi:10.18653/v1/2024.gebnlp-1.7 2024

[12] [12]

Sarvam translate

Sarvam AI . Sarvam translate. https://www.sarvam.ai/blogs/sarvam-translate, 2024. Accessed: March 2026

2024

[13] [13]

Saunders, R

D. Saunders, R. Sallis, and B. Byrne. Neural machine translation doesn't translate gender coreference right unless you make it. In Proceedings of the Second Workshop on Gender Bias in Natural Language Processing, pages 35--43. Association for Computational Linguistics, 2020

2020

[14] [14]

Saunders, R

blue D. Saunders, R. Sallis, and B. Byrne. First the Worst: Finding Better Gender Translations During Beam Search . In Findings of the Association for Computational Linguistics: ACL 2022 . Association for Computational Linguistics , 2022

2022

[15] [15]

Savoldi, M

B. Savoldi, M. Gaido, L. Bentivogli, M. Negri, and M. Turchi. Gender bias in machine translation. Transactions of the Association for Computational Linguistics, 9: 0 845--874, 2021. doi:10.1162/tacl_a_00401

work page doi:10.1162/tacl_a_00401 2021

[16] [16]

P. Singh. Gender inflected or bias inflicted: On using grammatical gender cues for bias evaluation in machine translation. In Proceedings of the 13th International Joint Conference on Natural Language Processing and the 3rd Conference of the Asia-Pacific Chapter of the ACL: Student Research Workshop, pages 17--23, 2023. doi:10.18653/v1/2023.ijcnlp-srw.3

work page doi:10.18653/v1/2023.ijcnlp-srw.3 2023

[17] [17]

Singh, M

P. Singh, M. Patidar, and L. Vig. Translating across cultures: LLM s for intralingual cultural adaptation. In Proceedings of the 28th Conference on Computational Natural Language Learning, pages 400--418, Miami, FL, USA, 2024. Association for Computational Linguistics. doi:10.18653/v1/2024.conll-1.30. URL https://aclanthology.org/2024.conll-1.30/

work page doi:10.18653/v1/2024.conll-1.30 2024

[18] [18]

Stafanovi c s, T

A. Stafanovi c s, T. Bergmanis, and M. Pinnis. Mitigating gender bias in machine translation with target gender annotations. In Proceedings of the Fifth Conference on Machine Translation, pages 629--638, Online, 2020. Association for Computational Linguistics. doi:10.18653/v1/2020.wmt-1.73. URL https://aclanthology.org/2020.wmt-1.73/

work page doi:10.18653/v1/2020.wmt-1.73 2020

[19] [19]

Stanovsky, N

G. Stanovsky, N. A. Smith, and L. Zettlemoyer. Evaluating gender bias in machine translation. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 1679--1684. Association for Computational Linguistics, 2019. doi:10.18653/v1/P19-1164

work page doi:10.18653/v1/p19-1164 2019

[20] [20]

Tiedemann and S

J. Tiedemann and S. Thottingal. OPUS-MT : Building open translation services for the world. In Proceedings of the 22nd Annual Conference of the European Association for Machine Translation, pages 479--480, 2020

2020

[21] [21]

Vanmassenhove, C

blue E. Vanmassenhove, C. Hardmeier, and A. Way. Getting Gender Right in Neural Machine Translation . In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing . Association for Computational Linguistics , 2018

2018

[22] [22]

B. Yao, M. Jiang, T. Bobinac, D. Yang, and J. Hu. Benchmarking machine translation with cultural awareness. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 13078--13096, Miami, Florida, USA, 2024. Association for Computational Linguistics. doi:10.18653/v1/2024.findings-emnlp.765. URL https://aclanthology.org/2024.findings-e...

work page doi:10.18653/v1/2024.findings-emnlp.765 2024

[23] [23]

J. Zhao, T. Wang, M. Yatskar, V. Ordonez, and K.-W. Chang. Gender bias in coreference resolution: Evaluation and debiasing methods. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 15--20. Association for Computational Linguistics, 2018. doi:10.18653/v1...

work page doi:10.18653/v1/n18-2003 2018

[24] [24]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...