Pith. sign in

REVIEW 5 major objections 4 minor 70 references

GeNRe: A French Gender-Neutral Rewriting System Using Collective Nouns

T0 review · 5 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read GeNRe is the first French gender-neutral rewriting system built on collective nouns, and its rule-based version reports 3.81% WER and 99.05 cosine similarity on a 500-sentence test set.

desk verdict Useful first French CN-based neutralizer with a reusable dictionary, but the headline metrics rest on an author-written gold set; worth refereeing for the resource. read the letter →

arxiv 2505.23630 v1 pith:VBC35HSE submitted 2025-05-29 cs.CL

Enzo Doyen , Amalia Todirascu This is my paper
classification cs.CL
keywords gender-neutralrewritingFrenchNLPcollectivenounsmasculinegenericsgenderbiasmitigationrule-basedtextfine-tunedlanguagemodelsinclusive
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that French sentences using masculine generics can be automatically rewritten into neutral form by replacing masculine plural member nouns with gender-fixed collective nouns, for example 'les soldats' becoming 'l'armée'. It presents GeNRe, a rule-based system built on a manually constructed dictionary of 315 French collective nouns, and compares it with two fine-tuned sequence-to-sequence models and an instructed version of Claude 3 Opus. On a 500-sentence evaluation set drawn from French Wikipedia and Europarl, the rule-based system reports 3.81% word error rate and 99.05 cosine similarity against the authors' manually neutralized gold sentences, while Claude 3 Opus using the dictionary reaches 93.519 BLEU. This matters because French previously had automatic gender-fair rewriting that highlights feminine endings but no automatic neutralization system, and neutralization avoids the spelling changes and new punctuation that make inclusive writing contentious.

What carries the argument

The load-bearing object is the collective-noun-to-member-noun dictionary: 315 entries manually assembled from a published linguistic list, media sources, and a suffix-pattern collection for '-phonie' nouns. A collective noun in French has a fixed gender independent of the people it denotes, such as 'la police' or 'l'armée', so swapping a masculine plural member noun for its collective counterpart removes masculine marking and forces downstream grammatical agreement among determiners, adjectives, past participles, and pronouns to change. The rule-based system carries out this swap through a dependency parser to locate words syntactically tied to the member noun, a morphological inflector to re-inflect them, and additional corrections for past participles and object pronouns; the fine-tuned and instructed models are alternative vehicles for the same replacement operation.

What would settle it

Have a panel of native French speakers rate the 500 gold sentences and the corresponding GeNRe outputs for grammaticality and naturalness; if a substantial share of collective-noun rewrites are judged asemantic or rejected in favor of another formulation, the reported WER and BLEU scores would not demonstrate usable gender-neutral French.

Watch

Extended reading notes

Core claim

In the paper's own terms, GeNRe is the first French gender-neutral rewriting system that uses collective nouns as the neutralization mechanism. The central discovery is that replacing masculine plural member nouns with collective nouns of fixed grammatical gender, such as 'les lecteurs assidus' becoming 'le lectorat assidu', yields rewrites that match the authors' gold sentences closely enough that the rule-based system reaches 3.81% WER and 99.05 cosine similarity, and that an instructed language model prompted with the authors' dictionary reaches 93.519 BLEU. The paper also finds that fine-tuning T5 and M2M100 on rule-based output does not improve on the rule-based system, and that the intended neutralization strategy has a semantic failure mode: the paper's own error analysis reports that collective-noun rewrites are often judged asemantic, with low interannotator agreement on that judgment (26.85% on the Europarl portion).

Load-bearing premise

The evaluation assumes that each of the 500 test sentences has a single correct gender-neutral rewrite, namely the authors' manual gold, and that WER, BLEU, and cosine similarity against that gold measure how well the system neutralizes gender.

Editorial extensions

If this is right

  • French NLP pipelines gain a concrete way to reduce masculine generics in training data without altering word spellings or adding interpuncts, since the rewrite targets group-denoting nouns only.
  • The released 315-entry dictionary and the roughly 399,000 sentence pairs give other researchers a starting point for neutralization in French and, after dictionary construction, in similarly inflected languages.
  • The result that fine-tuned T5 and M2M100 do not beat the rule-based system suggests that for this task, explicit linguistic rules outperform learned sequence-to-sequence rewriting.
  • Prompting an instructed language model with the dictionary approaches the rule-based system's quality, opening a path to neutralization without hand-written grammatical rules.
  • Because the error analysis locates most failures in semantics, improving the system likely means adding context-aware selection of which collective noun to use, not better inflection.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A downstream test of the system's actual effect on bias, for example whether readers or language models exposed to GeNRe outputs show fewer male-biased interpretations than with the original masculine generics, would probe what the reported form-level scores do not.
  • The low interannotator agreement on semantic errors suggests that acceptable rewrites vary by speaker; a multiple-gold evaluation could change the relative ranking of the rule-based system and the dictionary-prompted language model.
  • Because the dictionary covers masculine plural member nouns, singular masculine generics such as 'un professeur' remain untouched; extending coverage to singular cases would broaden the system's reach.
  • The same dictionary-plus-neutralization pipeline could be tested for Spanish or Italian, where collective nouns also have fixed gender, by measuring whether the dependency-adjustment rules transfer without French-specific code.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 4 minor

Summary. The paper presents GeNRe, a French gender-neutral rewriting system that replaces masculine-plural member nouns with fixed-gender collective nouns (e.g., 'les lecteurs' -> 'le lectorat'). Three system variants are described: a rule-based system (RBS) using spaCy and a manually built collective-noun dictionary; two fine-tuned sequence-to-sequence models (T5-small and M2M100-418M) trained on RBS-generated pairs; and Claude 3 Opus prompted either directly (BASE), with the dictionary (DICT), or as a corrector of RBS output (CORR). Evaluation on 500 manually rewritten sentences from Wikipedia and Europarl reports WER, BLEU, and cosine similarity; the RBS achieves the best average WER (3.81%) and cosine similarity (99.05), while Claude-DICT achieves the best BLEU (93.519). The authors release the dictionary, filtered corpora, and code.

Significance. If the reported results are valid, this is the first French gender-neutral rewriting system based on collective nouns and provides a reproducible resource (dictionary, datasets, code) that should enable follow-up work. The RBS's dependency-detection improvement over spaCy (Table 2: average F1 0.7985 vs. 0.183) is a concrete, machine-checked contribution. However, the central evaluation claim is currently not established because the gold references are authored by the same team, the metric scores are reported without uncertainty estimates, and the paper's own error analysis shows that the semantic acceptability of collective-noun rewrites is highly subjective (Table 8, SEM agreement 26.85% on Europarl). These issues are fixable with additional experiments, and the underlying approach and released resources justify further review.

major comments (5)
  1. [Section 4.2, Table 3, Table 8] The 500-sentence gold set is the sole reference for WER/BLEU/cosine in Table 3. The gold sentences were written by the authors, and the paper's own data show that collective-noun rewrites are frequently judged asemantic: SEM has the lowest interannotator agreement (26.85% on Europarl, Table 8), and the Limitations state that many collective nouns are not actively used and can yield asemantic constructions. As a result, the reported scores measure agreement with one team's rewriting style rather than the acceptability or neutrality of the rewrites. Please add an external validation step (e.g., multiple native-speaker acceptability ratings on a sample) and/or report agreement against multiple gold references.
  2. [Table 3, Section 5] The headline comparisons (RBS 3.81% WER vs. T5 5.492% WER; Claude-DICT 93.519 BLEU vs. RBS 92.887 BLEU) are reported as point estimates without confidence intervals or significance tests. On 500 sentences these differences could be sampling noise. Please compute sentence-level paired bootstrap intervals or a paired significance test (e.g., Wilcoxon signed-rank) for WER and BLEU, and report them for each corpus and on average.
  3. [Section 4.2, Section 4.3.2] The paper does not state explicitly that the 500 evaluation sentences are disjoint from the 398,954 extracted sentences used to build the fine-tuning pairs. If the evaluation sentences are included in the fine-tuning data, the T5 and M2M100 results are invalid due to train/test leakage. Please specify the exact split and confirm that the 500 sentences were excluded from training and validation.
  4. [Section 4.3.2, Section 5] The fine-tuned models are trained exclusively on RBS-generated sentence pairs. Evaluating them against gold references that closely resemble RBS outputs biases the comparison in favor of the RBS and weakens the conclusion that fine-tuning does not improve over the RBS. A fairer comparison would fine-tune on gold/manual rewrites as well, or at least evaluate with human judgments of neutralization quality.
  5. [Abstract, Introduction] The abstract states that automatic gender neutralization 'has only been studied for English,' but Section 3 cites neutralization work for Italian (Piergentili et al., 2023a) and German (Lardelli and Gromann, 2023). The novelty claim should be scoped to French or to the collective-noun strategy to avoid an overstatement.
minor comments (4)
  1. [Section 5, Table 3] The paper uses cosine similarity as a complementary metric, but the abstract still headlines '99.05 cosine similarity'; consider removing it from the abstract or describing it as a secondary measure.
  2. [Table 3] Bold marks the best results per column; since RBS and Claude-DICT split the best scores, the text should state more clearly which system is considered the overall best and why.
  3. [Section 6, Table 8] The very low SEM interannotator agreement is reported but its implications for the validity of the error analysis are not discussed; a brief paragraph would help.
  4. [Appendix F, Table 6] There are a few typos (e.g., 'apos;s' in the translation, 'ssentences' in the Table 6 caption) and the example numbering could be aligned with the main text.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the central RBS and dictionary are evaluated against a separate, manually created gold standard, though that gold was written by the same team and is not independently validated.

full rationale

We walked the claimed derivation chain: the dictionary (Section 4.1) is a manually built resource; the rule-based system (Section 4.3.1) applies syntactic rules to replace member nouns with dictionary collective nouns; and the gold standard (Section 4.2) is a separate set of 500 sentences manually gender-neutralized by the authors. The headline metrics in Table 3 compare system outputs to that gold using WER, BLEU, and cosine similarity; none of these quantities is defined in terms of the system's own outputs, and no parameter is fitted to the gold. The fine-tuned models are trained on RBS-generated sentence pairs, but the paper does not claim they are independent predictors—it explicitly reports that they do not improve over the RBS, so this is a controlled comparison rather than a circular prediction. The Claude-DICT variant uses the same dictionary as the RBS, but dictionary lookup is the method under test, not a hidden input to the gold. The paper's own admission that cosine similarity is ill-suited (Section 5) and the low SEM interannotator agreement (Table 8) are validity and acceptability concerns about the gold standard, not circularity: they do not make any equation reduce to itself. No self-citations are load-bearing; the only cited prior work on collective nouns (Lecolle 2019) is an external linguistic resource. We therefore find no significant circularity, though the same-team gold standard is a real empirical weakness.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a manually constructed dictionary, an imperfect dependency and inflection pipeline, and a small author-written gold set; no new theoretical entities are introduced. The 315 dictionary entries are manually chosen resources, not fitted numeric parameters, and no hyperparameters were tuned against the evaluation set in a way that would count as free parameters in a derivation.

assumptions (4)
  • domain assumption French human collective nouns are gender-fixed and can replace masculine plural member nouns without changing core meaning.
    Section 4.1 and the generation component; supported by Lecolle (2019), but not validated for all 315 dictionary entries or all contexts in this paper.
  • ad hoc to paper The 500-sentence evaluation set is disjoint from the fine-tuning training pairs.
    Both sets are drawn from the same Wikipedia and Europarl corpora (Sections 4.2 and 4.3.2); the paper does not state that evaluation sentences were excluded from the 60,000-pair training sets.
  • domain assumption WER, BLEU, and cosine similarity against author-written gold rewrites are valid measures of gender-neutralization quality.
    Section 5; the paper cites Piergentili et al. (2023b) warning that semantic similarity may be ill-suited, and no native-speaker acceptability judgments are reported.
  • domain assumption The spaCy and inflecteur modules provide sufficiently reliable dependency and inflection analysis for the RBS.
    Section 4.3.1 reports inflecteur accuracy of 75.35% on 264 dependencies, so errors in upstream tooling propagate into RBS output.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GeNRe: A French Gender-Neutral Rewriting System Using Collective Nouns." pith.science (2026). https://pith.science/paper/VBC35HSE

@misc{pith2026250523630,
  author       = {Pith},
  title        = {Pith review of: GeNRe: A French Gender-Neutral Rewriting System Using Collective Nouns},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VBC35HSE}},
  note         = {Machine review of arXiv:2505.23630}
}
read the original abstract

A significant portion of the textual data used in the field of Natural Language Processing (NLP) exhibits gender biases, particularly due to the use of masculine generics (masculine words that are supposed to refer to mixed groups of men and women), which can perpetuate and amplify stereotypes. Gender rewriting, an NLP task that involves automatically detecting and replacing gendered forms with neutral or opposite forms (e.g., from masculine to feminine), can be employed to mitigate these biases. While such systems have been developed in a number of languages (English, Arabic, Portuguese, German, French), automatic use of gender neutralization techniques (as opposed to inclusive or gender-switching techniques) has only been studied for English. This paper presents GeNRe, the very first French gender-neutral rewriting system using collective nouns, which are gender-fixed in French. We introduce a rule-based system (RBS) tailored for the French language alongside two fine-tuned language models trained on data generated by our RBS. We also explore the use of instruct-based models to enhance the performance of our other systems and find that Claude 3 Opus combined with our dictionary achieves results close to our RBS. Through this contribution, we hope to promote the advancement of gender bias mitigation techniques in NLP for French.

Figures

Figures reproduced from arXiv: 2505.23630 by the authors.

Figure 1
Figure 1. Rule-based model replacement pipeline overview were chosen for their great text-to-text performance and their relatively small size, making the train￾ing process easier. Furthermore, as M2M100 had already been used by Veloso et al. (2023), we want to compare the results we can get for our specific task. Both models were fine-tuned using our two RBS-generated corpora (Wikipedia and Europarl) containing gender-neutral… view at source ↗
Figure 2
Figure 2. Error distribution for RBS, fine-tuned and instruct-based models The latter method is chosen because LLMs tend to be significantly more permissive in their genera￾tions, resulting in a broader range of error types that would complicate manual annotation. We define four high-level categories of errors, as illustrated in [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

70 extracted references · 39 canonical work pages

  1. [1]

    Bashar Alhafni, Nizar Habash, and Houda Bouamor. 2022 a . https://doi.org/10.18653/v1/2022.naacl-main.46 User- Centric Gender Rewriting . In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics : Human Language Technologies , pages 618--631, Seattle, United States. Association for Computational ...

    work page doi:10.18653/v1/2022.naacl-main.46 2022

  2. [2]

    The Shared Task on Gender Rewriting

    Bashar Alhafni, Nizar Habash, Houda Bouamor, Ossama Obeid, Sultan Alrowili, Daliyah Alzeer, Khawlah M. Alshanqiti, Ahmed ElBakry, Muhammad ElNokrashy, Mohamed Gabr, Abderrahmane Issam, Abdelrahim Qaddoumi, K. Vijay-Shanker , and Mahmoud Zyate. 2022 b . https://doi.org/10.48550/arXiv.2210.12410 The Shared Task on Gender Rewriting . http://arxiv.org/abs/221...

    work page Pith review arXiv doi:10.48550/arxiv.2210.12410 2022

  3. [3]

    Anthropic . 2024. The Claude 3 Model Family : Opus , Sonnet , Haiku

    work page 2024

  4. [4]

    Marion Bartl and Susan Leavy. 2024. https://doi.org/10.48550/arXiv.2407.04434 From ' Showgirls ' to ' Performers ': Fine-tuning with Gender-inclusive Language for Bias Reduction in LLMs . Preprint, arXiv:2407.04434

    arXiv doi:10.48550/arxiv.2407.04434 2024

  5. [5]

    Brent Berlin and Paul Kay. 1969. Basic Color Terms : Their Universality and Evolution . University of California Press

    work page 1969

  6. [6]

    Friederike Braun, Sabine Sczesny, and Dagmar Stahlberg. 2005. https://doi.org/10.1515/comm.2005.30.1.1 Cognitive Effects of Masculine Generics in German : An Overview of Empirical Findings . Communications, 30(1):1--21

    work page doi:10.1515/comm.2005.30.1.1 2005

  7. [7]

    Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss , Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, and 12 others. 2020. https://arxiv.org/abs/2005.14165 La...

    arXiv 2020

  8. [8]

    Heather Burnett and C \'e line Pozniak. 2021. https://doi.org/10.1111/josl.12489 Political dimensions of gender inclusive writing in Parisian universities . Journal of Sociolinguistics, 25(5):808--831

    work page doi:10.1111/josl.12489 2021

Show all 70 references
  1. [9]

    Adrien Chuttarsing. 2021. Inflecteur

    2021

  2. [10]

    DeepSeek-AI , Aixin Liu, Bei Feng, Bing Xue, Bingxuan Wang, Bochao Wu, Chengda Lu, Chenggang Zhao, Chengqi Deng, Chenyu Zhang, Chong Ruan, Damai Dai, Daya Guo, Dejian Yang, Deli Chen, Dongjie Ji, Erhang Li, Fangyun Lin, Fucong Dai, and 181 others. 2024. https://doi.org/10.4855...

    2024 doi arXiv

  3. [11]

    Fanny Ducel, Aur \'e lie N \'e v \'e ol, and Kar \"e n Fort. 2024. \'E valuation automatique des biais de genre dans des mod \`e les de langue auto-r \'e gressifs . TALN 2024

    2024

  4. [12]

    Angela Fan, Shruti Bhosale, Holger Schwenk, Zhiyi Ma, Ahmed El-Kishky , Siddharth Goyal, Mandeep Baines, Onur Celebi, Guillaume Wenzek, Vishrav Chaudhary, Naman Goyal, Tom Birch, Vitaliy Liptchinsky, Sergey Edunov, Edouard Grave, Michael Auli, and Armand Joulin. 2020. https://...

    2020 arXiv

  5. [13]

    Nelly Flaux. 1999. https://doi.org/10.5169/SEALS-400007 \`A propos des noms collectifs . Revue de linguistique romane, (63):471--502

    1999 doi

  6. [14]

    Gygax, and Elisabeth A

    Ute Gabriel, Pascal M. Gygax, and Elisabeth A. Kuhn. 2018. https://doi.org/10.1177/1368430218771742 Neutralising linguistic sexism: Promising but cumbersome? Group Processes & Intergroup Relations, 21(5):844--858

    2018 doi

  7. [15]

    graelo . 2023. Graelo/wikipedia

    2023

  8. [16]

    Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle , Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, Amy Yang, Angela Fan, Anirudh Goyal, Anthony Hartshorn, Aobo Yang, Archi Mitra, Archie Sravankumar, Artem Korenev, Ar...

    2024 doi arXiv

  9. [17]

    Jiawei Gu, Xuhui Jiang, Zhichao Shi, Hexiang Tan, Xuehao Zhai, Chengjin Xu, Wei Li, Yinghan Shen, Shengjie Ma, Honghao Liu, Saizhuo Wang, Kun Zhang, Yuanzhuo Wang, Wen Gao, Lionel Ni, and Jian Guo. 2025. https://arxiv.org/abs/2411.15594 A survey on llm-as-a-judge . Preprint, a...

    2025 arXiv

  10. [18]

    Pascal Gygax, Ute Gabriel, Arik L \'e vy, Eva Pool, Marjorie Grivel, and Elena Pedrazzini. 2012. https://doi.org/10.1080/20445911.2011.642858 The masculine form and its competing interpretations in French : When linking grammatically masculine role names to female referents is...

    2012

  11. [19]

    Pascal Gygax, Ute Gabriel, Oriane Sarrasin, Jane Oakhill, and Alan Garnham. 2008. https://doi.org/10.1080/01690960701702035 Generically intended, but specifically interpreted: When beauticians, musicians, and mechanics are all men . Language and Cognitive Processes, 23(3):464--485

    2008 doi

  12. [20]

    Pascal Mark Gygax, Lucie Schoenhals, Arik L \'e vy, Patrick Luethold, and Ute Gabriel. 2019. https://doi.org/10.3389/fpsyg.2019.01225 Exploring the Onset of a Male-Biased Interpretation of Masculine Generics Among French Speaking Kindergarten Children . Frontiers in Psychology...

    2019

  13. [21]

    Nizar Habash, Houda Bouamor, and Christine Chung. 2019. https://doi.org/10.18653/v1/W19-3822 Automatic Gender Identification and Reinflection in Arabic . In Proceedings of the First Workshop on Gender Bias in Natural Language Processing , pages 155--165, Florence, Italy. Assoc...

    2019 doi

  14. [22]

    Harris Interactive . 2017. L' \'e criture inclusive : La population fran c aise conna \^i t-elle l' \'e criture inclusive ? Quelle opinion en a-t-elle ? Technical report

    2017

  15. [23]

    Zexue He, Bodhisattwa Prasad Majumder, and Julian McAuley. 2021. https://doi.org/10.48550/arXiv.2109.11708 Detect and Perturb : Neutral Rewriting of Biased and Sensitive Text via Gradient-based Decoding . http://arxiv.org/abs/2109.11708. Preprint, arXiv:2109.11708

    2021 doi arXiv

  16. [24]

    Jacobson and William R

    Marsha B. Jacobson and William R. Insko. 1985. https://doi.org/10.1007/BF00287456 Use of Nonsexist Pronouns as a Function of One 's Feminist Orientation . Sex Roles, 13(1-2):1--7

    1985 doi

  17. [25]

    Fanny Jourdan, Yannick Chevalier, and Cécile Favre. 2025. https://doi.org/10.48550/arXiv.2504.15941 FairTranslate : An English-French Dataset for Gender Bias Evaluation in Machine Translation by Overcoming Gender Binarity . Preprint, arXiv:2504.15941

    2025 doi arXiv

  18. [26]

    McDaniel

    Paul Kay and Chad K. McDaniel. 1978. The Linguistic Significance of the Meanings of Basic Color Terms . Language, 54(3):610--646

    1978

  19. [27]

    Adam Kilgarriff, Vít Baisa, Jan Bušta, Miloš Jakubíček, Vojtěch Kovář, Jan Michelfeit, Pavel Rychlý, and Vít Suchomel. 2014. The sketch engine: ten years on. Lexicography, 1:7--36

    2014

  20. [28]

    Philipp Koehn. 2005. Europarl: A Parallel Corpus for Statistical Machine Translation

    2005

  21. [29]

    Kuhn, and Sabine Sczesny

    Sara Koeser, Elisabeth A. Kuhn, and Sabine Sczesny. 2015. https://doi.org/10.1177/0261927X14561119 Just reading? how gender-fair language triggers readers' use of gender-fair forms. Journal of Language and Social Psychology, 34(3):343--357

    2015 doi

  22. [30]

    Marlene Kollmayer, Andreas Pfaffel, Barbara Schober, and Laura Brandt. 2018. https://doi.org/10.3389/fpsyg.2018.00985 Breaking away from the male stereotype of a specialist: Gendered language affects performance in a thinking task . Frontiers in Psychology, Volume 9 - 2018

    2018

  23. [31]

    Hadas Kotek, Rikker Dockum, and David Q. Sun. 2023. https://doi.org/10.1145/3582269.3615599 Gender bias and stereotypes in Large Language Models . In Proceedings of The ACM Collective Intelligence Conference , pages 12--24

    2023

  24. [32]

    Marie Lammert. 2010. S \'e mantique et Cognition : Les Noms Collectifs . Droz, Gen \`e ve

    2010

  25. [33]

    Marie Lammert and Michelle Lecolle. 2014. Les noms collectifs en fran c ais, une vue d'ensemble. Cahiers de lexicologie, (105):203--222

    2014

  26. [34]

    Manuel Lardelli and Dagmar Gromann. 2023. https://doi.org/10.5281/ZENODO.7898328 Gender- Fair Post-Editing : A Case Study Beyond the Binary

    2023 doi

  27. [35]

    Michelle Lecolle. 2019. Les Noms Collectifs Humains En Fran c ais. Enjeux S \'e mantiques, Lexicaux et Discursifs . Lambert-Lucas, Universit \'e de Lorraine

    2019

  28. [36]

    Paul Lerner and Cyril Grouin. 2024. INCLURE : A Dataset and Toolkit for Inclusive French Translation

    2024

  29. [37]

    Pengfei Liu, Weizhe Yuan, Jinlan Fu, Zhengbao Jiang, Hiroaki Hayashi, and Graham Neubig. 2021. https://arxiv.org/abs/2107.13586 Pre- Train , Prompt , and Predict : A Systematic Survey of Prompting Methods in Natural Language Processing . http://arxiv.org/abs/2107.13586. Prepri...

    2021 arXiv

  30. [38]

    Kaiji Lu, Piotr Mardziel, Fangjing Wu, Preetam Amancharla, and Anupam Datta. 2020. https://doi.org/10.1007/978-3-030-62077-6_14 Gender Bias in Neural Natural Language Processing . In Vivek Nigam, Tajana Ban Kirigin, Carolyn Talcott, Joshua Guttman, Stepan Kuznetsov, Boon Thau ...

    2020 doi

  31. [39]

    Louis Martin, Benjamin Muller, Pedro Javier Ortiz Su \'a rez, Yoann Dupont, Laurent Romary, \'E ric Villemonte de la Clergerie, Djam \'e Seddah, and Beno \^i t Sagot. 2020. https://doi.org/10.18653/v1/2020.acl-main.645 CamemBERT : A Tasty French Language Model . In Proceedings...

    2020 doi

  32. [40]

    Ines Montani, Matthew Honnibal, Adriane Boyd, Sofie Van Landeghem, and Henning Peters. 2024. https://doi.org/10.5281/ZENODO.1212303 spaCy : Industrial-strength Natural Language Processing in Python . Zenodo

    2024 doi

  33. [41]

    Mara Nunziatini and Sara Diego. 2024. https://aclanthology.org/2024.eamt-1.48/ Implementing gender-inclusivity in MT output using automatic post-editing with LLM s . In Proceedings of the 25th Annual Conference of the European Association for Machine Translation (Volume 1), pa...

    2024

  34. [42]

    OpenAI . 2024. GPT-4o mini: Advancing cost-efficient intelligence. https://openai.com/index/gpt-4o-mini-advancing-cost-efficient-intelligence/

    2024

  35. [43]

    Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L. Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul Christiano, Jan Leike, and...

    2022 arXiv

  36. [44]

    Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. https://doi.org/10.3115/1073083.1073135 BLEU : A Method for Automatic Evaluation of Machine Translation . In Proceedings of the 40th Annual Meeting on Association for Computational Linguistics - ACL '02 , Phila...

    2002

  37. [45]

    Andrea Piergentili, Dennis Fucci, Beatrice Savoldi, Luisa Bentivogli, and Matteo Negri. 2023 a . https://arxiv.org/abs/2301.10075 From Inclusive Language to Gender-Neutral Machine Translation . http://arxiv.org/abs/2301.10075. Preprint, arXiv:2301.10075

    2023 arXiv

  38. [46]

    Andrea Piergentili, Beatrice Savoldi, Dennis Fucci, Matteo Negri, and Luisa Bentivogli. 2023 b . https://arxiv.org/abs/2310.05294 Hi guys or hi folks? benchmarking gender-neutral machine translation with the gente corpus . Preprint, arXiv:2310.05294

    2023 arXiv

  39. [47]

    David Pomerenke. 2022. https://arxiv.org/abs/2212.02564 INCLUSIFY : A Benchmark and a Model for Gender-Inclusive German . http://arxiv.org/abs/2212.02564. Preprint, arXiv:2212.02564

    2022 arXiv

  40. [48]

    Matt Post. 2018. https://www.aclweb.org/anthology/W18-6319 A call for clarity in reporting BLEU scores . In Proceedings of the Third Conference on Machine Translation: Research Papers, pages 186--191, Belgium, Brussels. Association for Computational Linguistics

    2018

  41. [49]

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. 2020. https://arxiv.org/abs/1910.10683 Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer . http://arxiv.org/abs/1910.10...

    2020 arXiv

  42. [50]

    Nils Reimers and Iryna Gurevych. 2019. http://arxiv.org/abs/1908.10084 Sentence-bert: Sentence embeddings using siamese bert-networks . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics

    2019 arXiv

  43. [51]

    C \'e lia Richy and Heather Burnett. 2021. https://doi.org/10.4000/glad.2839 D \'e m \^e ler les effets des st \'e r \'e otypes et le genre grammatical dans le biais masculin : Une approche exp \'e rimentale . GLAD!, (10)

    2021 doi

  44. [52]

    Patrick Rothermund and Fritz Strack. 2024. https://doi.org/10.1177/0261927X241237739 Reminding May Not Be Enough : Overcoming the Male Dominance of the Generic Masculine . Journal of Language and Social Psychology, 43(4):468--485

    2024 doi

  45. [53]

    Beatrice Savoldi, Marco Gaido, Luisa Bentivogli, Matteo Negri, and Marco Turchi. 2021. https://doi.org/10.1162/tacl_a_00401 Gender Bias in Machine Translation . Transactions of the Association for Computational Linguistics, 9:845--874

    2021 doi

  46. [54]

    Sabine Sczesny, Magda Formanowicz, and Franziska Moser. 2016. https://doi.org/10.3389/fpsyg.2016.00025 Can Gender-Fair Language Reduce Gender Stereotyping and Discrimination ? Frontiers in Psychology, 7

    2016

  47. [55]

    Elsa Spinelli, Jean-Pierre Chevrot, and L \'e o Varnet. 2023. https://doi.org/10.3389/fpsyg.2023.1256779 Neutral is not fair enough: Testing the efficiency of different language gender-fair strategies . Frontiers in Psychology, 14:1256779

    2023

  48. [56]

    Dagmar Stahlberg, Sabine Sczesny, and Friederike Braun. 2001. https://doi.org/10.1177/0261927X01020004004 Name Your Favorite Musician : Effects of Masculine Generics and of their Alternatives in German . Journal of Language and Social Psychology, 20(4):464--469

    2001 doi

  49. [57]

    Karolina Stanczak and Isabelle Augenstein. 2021. https://arxiv.org/abs/2112.14168 A Survey on Gender Bias in Natural Language Processing . http://arxiv.org/abs/2112.14168. Preprint, arXiv:2112.14168

    2021 arXiv

  50. [58]

    Tony Sun, Andrew Gaut, Shirlyn Tang, Yuxin Huang, Mai ElSherief, Jieyu Zhao, Diba Mirza, Elizabeth Belding, Kai-Wei Chang, and William Yang Wang. 2019. https://doi.org/10.18653/v1/P19-1159 Mitigating Gender Bias in Natural Language Processing : Literature Review . Proceedings ...

    2019 doi

  51. [59]

    Tony Sun, Kellie Webster, Apu Shah, William Yang Wang, and Melvin Johnson. 2021. https://arxiv.org/abs/2102.06788 They, Them , Theirs : Rewriting with Gender-Neutral English . Preprint, arXiv:2102.06788

    2021 arXiv

  52. [60]

    Julia Tibblin, Jonas Granfeldt, Joost Van De Weijer, and Pascal Gygax. 2023. https://doi.org/10.5070/G60111267 The male bias can be attenuated in reading: On the resolution of anaphoric expressions following gender-fair forms in French . Glossa Psycholinguistics, 2(1)

    2023 doi

  53. [61]

    Eva Vanmassenhove. 2024. https://doi.org/10.48550/arXiv.2401.10016 Gender Bias in Machine Translation and The Era of Large Language Models . Preprint, arXiv:2401.10016

    2024 doi arXiv

  54. [62]

    Eva Vanmassenhove, Chris Emmery, and Dimitar Shterionov. 2021. https://doi.org/10.48550/arXiv.2109.06105 NeuTral Rewriter : A Rule-Based and Neural Approach to Automatic Rewriting into Gender-Neutral Alternatives . http://arxiv.org/abs/2109.06105. Preprint, arXiv:2109.06105

    2021 doi arXiv

  55. [63]

    Leonor Veloso, Luisa Coheur, and Rui Ribeiro. 2023. https://doi.org/10.18653/v1/2023.findings-emnlp.585 A Rewriting Approach for Gender Inclusivity in Portuguese . In Findings of the Association for Computational Linguistics : EMNLP 2023 , pages 8747--8759, Singapore. Associat...

    2023 doi

  56. [64]

    Jean-Philippe Watbled. 2012. Linguistique du genre. L'Harmattan, pages 167--179

    2012

  57. [65]

    Guillaume Wisniewski, Lichao Zhu, Nicolas Ballier, and Fran c ois Yvon. 2021. https://doi.org/10.18653/v1/2021.blackboxnlp-1.24 Screening Gender Transfer in Neural Machine Translation . In Proceedings of the Fourth BlackboxNLP Workshop on Analyzing and Interpreting Neural Netw...

    2021 doi

  58. [66]

    Hualin Xiao, Brent Strickland, and Sharon Peperkamp. 2023. https://doi.org/10.1177/0261927X221084643 How fair is gender-fair language? insights from gender ratio estimations in french . Journal of Language and Social Psychology, 42(1):82--106

    2023 doi

  59. [67]

    Jieyu Zhao, Tianlu Wang, Mark Yatskar, Vicente Ordonez, and Kai-Wei Chang. 2018. https://doi.org/10.48550/arXiv.1804.06876 Gender Bias in Coreference Resolution : Evaluation and Debiasing Methods . Preprint, arXiv:1804.06876

    2018 doi arXiv

  60. [68]

    Xing, Hao Zhang, Joseph E

    Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric P. Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica. 2023. https://arxiv.org/abs/2306.05685 Judging llm-as-a-judge with mt-bench and chatbot arena . P...

    2023 arXiv

  61. [69]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  62. [70]

    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 gl...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.