Pith. sign in

REVIEW 4 major objections 5 minor 42 references

Addressing speaker gender bias in large scale speech translation systems

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Fine-tuning a large speech-translation model on LLM-corrected gendered data lifts feminine accuracy on MuST-SHE from about 10% to above 84% without hurting BLEU.

desk verdict A practical, large-scale fix for a real bias problem, with a convincing external benchmark result and a soft spot in the Auto-mode mechanism. read the letter →

arxiv 2501.05989 v1 pith:T5DZGXUF submitted 2025-01-10 cs.CL cs.AI

Shubham Bansal , Vikas Joshi , Harveen Chadha , Rupeshkumar Mehta , Jinyu Li This is my paper · ORCID
classification cs.CLcs.AI
keywords speechtranslationgenderbiasgenderedaccuracyLLMreformulationfine-tuningMuST-SHEstreamingtransducerrepresentationloss
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 the masculine bias of large direct speech-translation models can be largely removed by a two-step recipe: use a large language model to rewrite a carefully selected subset of training translations into distinct masculine and feminine forms matched to the speaker's gender, then fine-tune the ST model on that corrected data. On the public MuST-SHE test set, the fine-tuned model raises Category-1 feminine gendered-translation accuracy from 10.12% to 87.05% for English–Spanish and from 8.81% to 84.72% for English–Italian, while BLEU stays essentially unchanged. The paper also adds a three-mode model that can produce masculine, feminine, or audio-inferred translations on demand, and a gender-representation loss that helps the encoder extract gender information from speech. The significance is that gender-correct translation can be obtained directly from audio at scale, without requiring the user to supply gender labels at inference time.

What carries the argument

The central mechanism is a data-reformulation and fine-tuning loop. First, training data is filtered to utterances containing English first-person pronouns, which cover 97.8% of Category-1 gendered cases in the MuST-SHE dev set; a balanced 2M-utterance subset is then passed through GPT-4 with few-shot and chain-of-thought prompting to produce separate masculine and feminine target translations for each utterance. Fine-tuning STBase on these gender-debiased targets along with gender-neutral data yields the 1-mode model, while the 3-mode model prepends a mode token such as <Lang_Auto>, <Lang_Masc>, or <Lang_Femi> to the decoder target so translation mode can be chosen at inference. A gender-representation loss adds a small classifier on encoder output that is trained to predict the speaker's gender label, nudging the encoder to preserve acoustic gender cues. These pieces together let a transducer-based streaming ST model generate gender-correct forms without explicit gender input at inference.

What would settle it

Run the 1-mode fine-tuned Auto model on a held-out set of female speakers from dialects and recording conditions not represented in the 75,000-hour corpus and measure Category-1 feminine GTA on MuST-SHE; if accuracy drops back toward the 10% baseline because the model relied on acoustic correlates of the training corpus rather than speaker gender, the central claim fails. Corruption or random shuffling of the speaker-gender labels in the fine-tuning subset should also erase the Auto-mode gains if the gender mapping is real.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that a large-scale streaming direct ST model (STBase) can be adapted to respect speaker gender by fine-tuning on roughly 2 million LLM-reformulated utterances rather than retraining on hundreds of thousands of hours. The key quantitative claim, reported in Table 2, is that Category-1 feminine gendered-translation accuracy (GTA) on MuST-SHE rises from single digits to the mid-80s for both English–Spanish and English–Italian, with masculine accuracy, BLEU, and Category-2 accuracy all holding roughly steady. A companion three-mode fine-tuned model uses language-mode start tokens so the same model either infers gender from audio (Auto) or follows an explicit Masculine/Feminine choice, with average GTA above 87.5% across all speakers. The paper further claims that the gender-representation loss improves accuracy when neutral data is mixed in, and that this method outperforms large-scale ST systems such as Seamless M4T and Canary on feminine forms.

Load-bearing premise

The weakest load-bearing premise is that the Audio mode's speaker-gender labels in the 75,000-hour internal corpus are accurate enough for the encoder to learn a transferable audio-to-gender mapping; the paper filters to utterances with labels but reports no audit of label accuracy.

Editorial extensions

If this is right

  • Female speakers' translations improve by roughly 77 absolute percentage points on Category-1 MuST-SHE, so direct ST models no longer default to masculine forms for self-referential sentences.
  • The LLM-reformulation step only needs a 2M-utterance subset, so gender debiasing is affordable for a 75,000-hour model without degrading BLEU.
  • The three-mode model gives users control: Auto infers gender from speech, while Masculine/Feminine modes force the form regardless of audio, covering use cases where gender is predefined or should not be inferred.
  • Because the prompting method is language-agnostic, the same recipe should carry to other grammatical-gender target languages beyond Spanish and Italian.

Reading between the lines

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

  • An untested risk is that the Auto mode's gains depend on the internal corpus's speaker-gender labels; if those labels are noisy or correlated with acoustic factors other than gender, the accuracy may not transfer to new dialects, recording conditions, or speakers, so a label-accuracy audit and a held-out speaker test would be the natural next check.
  • The paper only evaluates GPT-4 as the reformulator; a cheaper or open-weight LLM with comparable few-shot gender rewriting could make the pipeline reproducible outside large industrial setups, but that extension is not demonstrated.
  • The gender scheme is binary, so Masculine/Feminine modes do not address non-binary or gender-indeterminate speakers; the paper lists that as future work.
  • Since 97.8% of Category-1 cases are signaled by first-person pronouns, a text-only filter plus LLM rewrite is a cheap, generalizable recipe; one could test whether the same 2M-sample fine-tuning works on other MuST-SHE language pairs such as English–French or English–German.
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

4 major / 5 minor

Summary. The paper addresses speaker-gender bias in large-scale direct speech translation (ST). The authors first select English utterances containing first-person pronouns from a 75,000-hour internal corpus, use GPT-4 with few-shot and chain-of-thought prompting to produce feminine and masculine reformulations of the translations, and then fine-tune a proprietary transformer-transducer ST model (STBase) on the reformulated targets mixed with gender-neutral data. Two fine-tuning variants are proposed: a 1-mode model trained on gender-debiased targets only, and a 3-mode model that adds explicit Masculine, Feminine, and Auto modes via language/mode start-of-sentence tokens. A gender-representation (GR) loss is added to encourage the encoder to predict speaker gender from audio. On the MuST-SHE test set, the 1-mode model raises Category-1 feminine GTA from 10.12% to 87.05% (EN-ES) and from 8.81% to 84.72% (EN-IT), with BLEU roughly unchanged, and the 3-mode model reports average GTA above 87.5%. The paper also compares with Seamless M4T and Canary and reports large advantages on the same metric.

Significance. If the results hold, the paper makes a strong practical contribution: it shows that a modest amount of LLM-reformulated data (2M utterances) can be used to fine-tune a large direct ST model, removing the masculine-default bias that is otherwise inherited from MT-generated training data, while preserving BLEU. The evaluation on the external MuST-SHE test set, the use of the dev split only for hyperparameter tuning, the consistent gains in two target languages, and the prompt-quality check on MuST-SHE dev are all notable strengths. The three-mode design (Auto/Masculine/Feminine) is a sensible response to the ethical concern that gender should not always be inferred from voice. However, the paper does not release code, data, or the STBase model, and the training corpus is proprietary; combined with the absence of uncertainty quantification and the unexamined possibility of train/test overlap, this currently limits the strength of the causal claims about learning a transferable audio-to-gender mapping.

major comments (4)
  1. [§4.3, Table 2] The central claim of a roughly 77-percentage-point gain in Category-1 feminine GTA is presented as a single point estimate with no confidence intervals, bootstrap errors, or significance tests in Tables 2–4. On a hand-annotated subset like MuST-SHE, the number of gendered tokens per category can be small, and the per-word GTA metric is categorical, so the reader cannot tell whether the reported differences are stable or whether the comparison to Seamless M4T and Canary is statistically meaningful. Please report percentile-bootstrap confidence intervals over utterances for GTA and BLEU, and a paired significance test for the Table 4 comparisons.
  2. [§4.1.1] The training corpus is described only as an internal 75,000-hour corpus 'procured through different sources,' while MuST-SHE is built from TED talks. No train/test overlap or speaker-overlap analysis is reported. If any TED or MuST-C audio appears in the 75,000 hours, the Category-1 GTA gains in Table 2 could be inflated by speaker or recording memorization rather than by a generalizable gender mapping. Please report (i) whether MuST-SHE/MuST-C speakers or audio appear in the training data, (ii) the same GTA figures after excluding any overlapping speakers or recordings, and (iii) ideally, an evaluation on an out-of-domain gender-annotated ST set.
  3. [§3.2 and §3.3] The abstract and Section 3.2 claim that in Auto mode the model generates gender-specific translations 'directly from audio cues,' but the only evidence is the MuST-SHE GTA improvement. There is no diagnostic showing that the encoder's gender prediction trained by the GR loss generalizes to held-out voices or that it does not exploit non-gender acoustic correlates (microphone, age, prosody, recording conditions) that align with gender in TED talks. I request an analysis of the GR head's accuracy on held-out speakers, an ablation with voice-morphing or cross-corpus data where gender is swapped or controlled, and a report of Auto-mode GTA broken down by speaker or recording condition if such metadata exist.
  4. [§3.1.1 and §4.3.2] The data-centric method depends on (a) GPT-4 reformulation quality and (b) the speaker-gender labels used for sampling and for the GR loss, yet neither is audited at scale. Section 4.3.2 reports 94%/93% prompt accuracy on the MuST-SHE dev set, which is not a sample from the 2M-utterance training distribution, and Section 3.1.1 says utterances without labels are ignored without reporting label accuracy. Please provide a human or automated audit of a few hundred to a thousand reformulated training pairs (speaker-gender correctness, referent-gender preservation, translation adequacy) and an estimate of label accuracy on the fine-tuning subset.
minor comments (5)
  1. [Abstract] The phrase 'absolute 70% improvement' should read '70 percentage points' or 'an absolute improvement of about 77 points'; the relative improvement over 10% GTA is much larger.
  2. [References] References [33] and [38] are the same BLEU paper and should be merged.
  3. [References, §1] Reference [29] (Llama 2) is cited together with [30] as support for GPT-4; the Llama 2 citation is not relevant here and should be removed or replaced with the GPT-4 technical report.
  4. [§3.3] The GR-loss equation has formatting issues (e.g., 'gu' and 'ot' should be g_u and o_t, 'sof tmax' is misspelled), and the dimensions of W_out and W_g are not defined; please state how the frame-level gender prediction is aggregated or used during fine-tuning.
  5. [§4.3] The text states that term coverage exceeds 80% for all models but does not report the numbers anywhere; please include term-coverage values in Table 2 or a separate table.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the gender-bias gains are measured on held-out MuST-SHE test data and do not reduce to fitted inputs or self-citations.

full rationale

After tracing the derivation chain, I find no circular step that reduces a prediction to the paper's own inputs by construction. The core result (Table 2) is a fine-tuned ST model evaluated on the publicly available MuST-SHE test set; neither the GPT-4 reformulated training targets nor the STBase weights are fitted to MuST-SHE test annotations. Hyperparameters θneut and α are tuned on MuST-SHE dev utterances (Sections 4.1.2 and 4.3.1), which is standard model selection and does not make the held-out test GTA numbers (87.05% EN-ES feminine, 84.72% EN-IT feminine) definitionally forced. The GPT-4 reformulation quality check in Section 4.3.2 is also on the dev set, and the final GTA is computed from model-generated translations, not from GPT-4 outputs on test sentences. The only self-citations (e.g., [7] for the transformer-transducer architecture, with co-author overlap) specify the architecture and are re-described in Section 4.2; they are not used to assert the gender-bias improvement. Skeptical concerns about whether Auto mode truly learns a transferable audio-to-gender mapping are external-validity and correctness risks, not circularity: the paper does not define its reported prediction metric in terms of the gender labels used for fine-tuning, and the evaluation is independent of the GPT-4 reformulation process.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on engineering assumptions rather than formal axioms: the first-person filter, audio-based gender learnability, GPT-4 target reliability, and train/test disjointness. The main hand-set quantities are theta_neut, alpha, and the 2M-utterance LLM reformulation budget. No new physical entities or invented theoretical constructs are introduced.

free parameters (3)
  • theta_neut = 0.2
    Ratio of gender-neutral data used during fine-tuning, tuned on the MuST-SHE dev split in Section 4.3.1, and it directly controls the trade-off between gender accuracy and BLEU preservation.
  • alpha = 0.1
    Weight of the gender representation loss in L_comb, tuned on the MuST-SHE dev split in Section 4.3.1, and it determines how strongly the encoder is pushed to retain gender information.
  • LLM reformulation budget = 2,000,000 utterances
    Number of first-person-containing utterances sampled for GPT-4 reformulation, chosen by hand in Section 3.1.1 to balance cost against adaptation quality; the paper does not sweep this value.
assumptions (4)
  • domain assumption Utterances with English first-person pronouns are a sufficient proxy for Category-1 gender-specific translation cases.
    This filter reduces the training subset to 19.4% of the original data in Section 3.1.1, supported by the observation that 97.8% of MuST-SHE dev Category-1 examples contain a first-person pronoun, but it is a corpus-specific heuristic.
  • domain assumption Speaker gender is recoverable from audio cues by the ST encoder.
    Auto mode and the gender representation loss in Section 3.3 assume the speech signal contains a learnable and generalizable gender cue; no demographic coverage, dialect robustness, or label-quality analysis is reported.
  • domain assumption GPT-4 reformulations are correct and preserve non-gender semantics.
    The fine-tuning pipeline in Sections 3.1.2 and 3.1.3 treats GPT-4 outputs as ground-truth targets, but accuracy is measured only on MuST-SHE dev (94% ES, 93% IT), not on the internal 2M-utterance training set.
  • domain assumption The internal 75k-hour training corpus does not overlap the MuST-SHE test set.
    No overlap analysis is provided in Section 4.1; if the internal corpus contains the same TED talks as MuST-SHE, the reported evaluation could be optimistic.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Addressing speaker gender bias in large scale speech translation systems." pith.science (2026). https://pith.science/paper/T5DZGXUF

@misc{pith2026250105989,
  author       = {Pith},
  title        = {Pith review of: Addressing speaker gender bias in large scale speech translation systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/T5DZGXUF}},
  note         = {Machine review of arXiv:2501.05989}
}
read the original abstract

This study addresses the issue of speaker gender bias in Speech Translation (ST) systems, which can lead to offensive and inaccurate translations. The masculine bias often found in large-scale ST systems is typically perpetuated through training data derived from Machine Translation (MT) systems. Our approach involves two key steps. First, we employ Large Language Models (LLMs) to rectify translations based on the speaker's gender in a cost-effective manner. Second, we fine-tune the ST model with the corrected data, enabling the model to generate gender-specific translations directly from audio cues, without the need for explicit gender input. Additionally, we propose a three-mode fine-tuned model for scenarios where the speaker's gender is either predefined or should not be inferred from speech cues. We demonstrate a 70% improvement in translations for female speakers compared to our baseline and other large-scale ST systems, such as Seamless M4T and Canary, on the MuST-SHE test set.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 30 canonical work pages

  1. [1]

    he” or “she,

    INTRODUCTION AND RELATED WORK Speech translation (ST) transforms spoken words to text in another language. They have applications in automatic video subtitling, dubbing, and facilitating cross-lingual communi- cation. Traditionally, ST followed a cascaded approach. First, a speech recognizer [1, 2] converted spoken words into text in the source language. ...

    arXiv 2025

  2. [2]

    I am a teacher

    TYPES OF GENDER BIAS Before delving into our methodology, it is crucial to outline the major categories of gender bias (as defined in latest MuST- SHE release), and specify the type of bias we aim to address: Category 1: In instances where the speaker’s gender can- not be deduced from the English text (e.g., “I am a teacher”), but can be inferred from aud...

    work page

  3. [3]

    Play a song

    METHOD Training data for direct speech translation models is often generated at scale by translating human-labeled speech recog- nition data using a MT model [35, 7]. This approach is cost- effective compared to obtaining human-labeled translations of speech data. As most MT models exhibit gender bias, the training data for direct ST models also inherits ...

    work page

  4. [4]

    dev” in the MuSTC-v1.0-SET column. The evaluation phase is limited to the remaining utterances not labeled as “dev

    EXPERIMENTAL RESULTS 4.1. Dataset 4.1.1. Training set Our research focuses on addressing gender bias in large- scale speech translation (ST) models. Due to the scarcity of such large amount of training datasets in the public domain, we have developed our baseline ST model using an inter- nal corpus. This corpus encompasses roughly 75,000 hours of speech d...

    work page

  5. [5]

    We also proposed three-mode fine-tuning to account for user prefer- ence and combined training with gender representation loss

    CONCLUSIONS Our study presented a novel approach to mitigating speaker gender bias in large-scale speech translation systems, lever- aging LLMs for gender-alternative translations. We also proposed three-mode fine-tuning to account for user prefer- ence and combined training with gender representation loss. Our methods have demonstrated a significant impr...

    work page

  6. [6]

    Transformer transducer: A streamable speech recog- nition model with transformer encoders and rnn-t loss,

    Qian Zhang, Han Lu, Hasim Sak, Anshuman Tripathi, Erik McDermott, Stephen Koo, and Shankar Kumar, “Transformer transducer: A streamable speech recog- nition model with transformer encoders and rnn-t loss,” in ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2020, pp. 7829–7833

    work page 2020

  7. [7]

    Im- proving rnn transducer modeling for end-to-end speech recognition,

    Jinyu Li, Rui Zhao, Hu Hu, and Yifan Gong, “Im- proving rnn transducer modeling for end-to-end speech recognition,” in 2019 IEEE Automatic Speech Recogni- tion and Understanding Workshop (ASRU). IEEE, 2019, pp. 114–121

    work page 2019

  8. [8]

    Neural machine translation by jointly learning to align and translate,

    Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Ben- gio, “Neural machine translation by jointly learning to align and translate,” arXiv preprint arXiv:1409.0473 , 2014

    arXiv 2014

Show all 42 references
  1. [9]

    Neural machine translation: A re- view,

    Felix Stahlberg, “Neural machine translation: A re- view,” Journal of Artificial Intelligence Research , vol. 69, pp. 343–418, 2020

    2020

  2. [10]

    End-to-end auto- matic speech translation of audiobooks,

    Alexandre B ´erard, Laurent Besacier, Ali Can Ko- cabiyikoglu, and Olivier Pietquin, “End-to-end auto- matic speech translation of audiobooks,” in 2018 IEEE International Conference on Acoustics, Speech and Sig- nal Processing (ICASSP). IEEE, 2018, pp. 6224–6228

    2018

  3. [11]

    On knowledge distillation for direct speech translation,

    Marco Gaido, Mattia A Di Gangi, Matteo Negri, and Marco Turchi, “On knowledge distillation for direct speech translation,” arXiv preprint arXiv:2012.04964 , 2020

    2012 arXiv

  4. [12]

    Large-Scale Streaming End-to-End Speech Translation with Neural Transducers,

    Jian Xue, Peidong Wang, Jinyu Li, Matt Post, and Yashesh Gaur, “Large-Scale Streaming End-to-End Speech Translation with Neural Transducers,” in Proc. Interspeech 2022, 2022, pp. 3263–3267

    2022

  5. [13]

    Re- cent advances in direct speech-to-text translation,

    Chen Xu, Rong Ye, Qianqian Dong, Chengqi Zhao, Tom Ko, Mingxuan Wang, Tong Xiao, and Jingbo Zhu, “Re- cent advances in direct speech-to-text translation,”arXiv preprint arXiv:2306.11646, 2023

    2023 arXiv

  6. [14]

    Literature survey: Spoken language trans- lation,

    Sanket Gandhare, Preethi Jyothi, and Pushpak Bhat- tacharyya, “Literature survey: Spoken language trans- lation,” Sanket SurveyPaper SPKMT. pdf, 2018

    2018

  7. [15]

    Cascade versus direct speech translation: Do the differences still make a difference?,

    Luisa Bentivogli, Mauro Cettolo, Marco Gaido, Alina Karakanta, Alberto Martinelli, Matteo Negri, and Marco Turchi, “Cascade versus direct speech translation: Do the differences still make a difference?,” arXiv preprint arXiv:2106.01045, 2021

    2021 arXiv

  8. [16]

    Cascade or direct speech translation? a case study,

    Thierry Etchegoyhen, Haritz Arzelus, Harritxu Gete, Aitor Alvarez, Iv ´an G Torre, Juan Manuel Mart ´ın- Do˜nas, Ander Gonz ´alez-Docasal, and Edson Benites Fernandez, “Cascade or direct speech translation? a case study,” Applied Sciences, vol. 12, no. 3, pp. 1097, 2022

    2022

  9. [17]

    These are not the stereotypes you are looking for: Bias and fairness in authorial gender attribution,

    Corina Koolen and Andreas van Cranenburgh, “These are not the stereotypes you are looking for: Bias and fairness in authorial gender attribution,” in Proceed- ings of the First Ethics in NLP workshop . Association for Computational Linguistics (ACL), 2017, pp. 12–22

    2017

  10. [18]

    Mitigat- ing gender bias in natural language processing: Litera- ture review,

    Tony Sun, Andrew Gaut, Shirlyn Tang, Yuxin Huang, Mai ElSherief, Jieyu Zhao, Diba Mirza, Elizabeth Beld- ing, Kai-Wei Chang, and William Yang Wang, “Mitigat- ing gender bias in natural language processing: Litera- ture review,” arXiv preprint arXiv:1906.08976, 2019

    1906 arXiv

  11. [19]

    Stere- oset: Measuring stereotypical bias in pretrained lan- guage models,

    Moin Nadeem, Anna Bethke, and Siva Reddy, “Stere- oset: Measuring stereotypical bias in pretrained lan- guage models,” arXiv preprint arXiv:2004.09456, 2020

    2004 arXiv

  12. [21]

    Gender in danger? evaluating speech transla- tion technology on the must-she corpus,

    Luisa Bentivogli, Beatrice Savoldi, Matteo Negri, Mat- tia Antonino Di Gangi, Roldano Cattoni, and Marco Turchi, “Gender in danger? evaluating speech transla- tion technology on the must-she corpus,” arXiv preprint arXiv:2006.05754, 2020

    2006 arXiv

  13. [22]

    Breeding gender-aware direct speech translation systems,

    Marco Gaido, Beatrice Savoldi, Luisa Bentivogli, Mat- teo Negri, and Marco Turchi, “Breeding gender-aware direct speech translation systems,” arXiv preprint arXiv:2012.04955, 2020

    2012 arXiv

  14. [23]

    Gender and dialect bias in youtube’s automatic captions,

    Rachael Tatman, “Gender and dialect bias in youtube’s automatic captions,” in Proceedings of the first ACL workshop on ethics in natural language processing , 2017, pp. 53–59

    2017

  15. [24]

    When does it hurt? inter- group permeability moderates the link between discrim- ination and self-esteem,

    David Bourguignon, Vincent Y Yzerbyt, Catia P Teix- eira, and Ginette Herman, “When does it hurt? inter- group permeability moderates the link between discrim- ination and self-esteem,” European Journal of Social Psychology, vol. 45, no. 1, pp. 3–9, 2015

    2015

  16. [25]

    Greville corbett, gender . (cambridge textbooks in linguistics.) cambridge: Cambridge univer- sity press, 1991. pp. xix + 363.,

    Anthony Aristar, “Greville corbett, gender . (cambridge textbooks in linguistics.) cambridge: Cambridge univer- sity press, 1991. pp. xix + 363.,” Journal of Linguistics, vol. 28, pp. 542 – 547, 1992

    1991

  17. [26]

    Reducing gender bias in neural machine translation as a domain adapta- tion problem,

    Danielle Saunders and Bill Byrne, “Reducing gender bias in neural machine translation as a domain adapta- tion problem,” in Proceedings of the 58th Annual Meet- ing of the Association for Computational Linguistics , Dan Jurafsky, Joyce Chai, Natalie Schluter, and Joel Tetreaul...

    2020

  18. [27]

    Neu- ral machine translation doesn’t translate gender corefer- ence right unless you make it,

    Danielle Saunders, Rosie Sallis, and Bill Byrne, “Neu- ral machine translation doesn’t translate gender corefer- ence right unless you make it,” in Proceedings of the Second Workshop on Gender Bias in Natural Language Processing, Marta R. Costa-juss`a, Christian Hardmeier, Wil...

    2020

  19. [28]

    Gender bias in machine translation,

    Beatrice Savoldi, Marco Gaido, Luisa Bentivogli, Mat- teo Negri, and Marco Turchi, “Gender bias in machine translation,” Transactions of the Association for Com- putational Linguistics, vol. 9, pp. 845–874, 2021

    2021

  20. [29]

    Gender aware spoken language translation applied to english-arabic,

    Hany Hassan Awadalla, Mostafa Elaraby, Ahmed Taw- fik, Mahmoud Khaled, and Aly Osama, “Gender aware spoken language translation applied to english-arabic,” in 2018 IEEE Proceedings of the Second International Conference on Natural Language and Speech Process- ing, February 2018

    2018

  21. [30]

    Test suites task: Evaluation of gender fair- ness in mt with must-she and ines,

    Beatrice Savoldi, Marco Gaido, Matteo Negri, and Luisa Bentivogli, “Test suites task: Evaluation of gender fair- ness in mt with must-she and ines,” arXiv preprint arXiv:2310.19345, 2023

    2023 arXiv

  22. [31]

    Seamless: Multilingual expres- sive and streaming speech translation,

    Lo ¨ıc Barrault, Yu-An Chung, Mariano Coria Megli- oli, David Dale, Ning Dong, Mark Duppenthaler, Paul- Ambroise Duquenne, Brian Ellis, Hady Elsahar, Justin Haaheim, et al., “Seamless: Multilingual expres- sive and streaming speech translation,” arXiv preprint arXiv:2312.05187, 2023

    2023 arXiv

  23. [32]

    Must-c: A multilingual corpus for end-to-end speech translation,

    Roldano Cattoni, Mattia Antonino Di Gangi, Luisa Ben- tivogli, Matteo Negri, and Marco Turchi, “Must-c: A multilingual corpus for end-to-end speech translation,” Computer Speech & Language , vol. 66, pp. 101155, 2021

    2021

  24. [33]

    Augmenting librispeech with french transla- tions: A multimodal corpus for direct speech translation evaluation,

    Ali Can Kocabiyikoglu, Laurent Besacier, and Olivier Kraif, “Augmenting librispeech with french transla- tions: A multimodal corpus for direct speech translation evaluation,” arXiv preprint arXiv:1802.03142, 2018

    2018 arXiv

  25. [34]

    Llama 2: Open foundation and fine-tuned chat mod- els,

    Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al., “Llama 2: Open foundation and fine-tuned chat mod- els,” arXiv preprint arXiv:2307.09288, 2023

    2023 arXiv

  26. [35]

    Gpt-4 technical report,

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al., “Gpt-4 technical report,” arXiv preprint arXiv:2303.08774, 2023

    2023 arXiv

  27. [36]

    Fairness-guided few- shot prompting for large language models,

    Huan Ma, Changqing Zhang, Yatao Bian, Lemao Liu, Zhirui Zhang, Peilin Zhao, Shu Zhang, Huazhu Fu, Qinghua Hu, and Bingzhe Wu, “Fairness-guided few- shot prompting for large language models,” arXiv preprint arXiv:2303.13217, 2023

    2023 arXiv

  28. [37]

    Chain-of-thought prompting elicits reasoning in large language models,

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al., “Chain-of-thought prompting elicits reasoning in large language models,” Advances in Neural Information Pro- cessing Systems, vol. 35, pp. 24824–24837, 2022

    2022

  29. [38]

    Bleu: a method for automatic evaluation of machine translation,

    Kishore Papineni, Salim Roukos, Todd Ward, and Wei- Jing Zhu, “Bleu: a method for automatic evaluation of machine translation,” in Proceedings of the 40th annual meeting of the Association for Computational Linguis- tics, 2002, pp. 311–318

    2002

  30. [39]

    Nvidia nemo canary model pushes the fron- tier of speech recognition and translation,

    “Nvidia nemo canary model pushes the fron- tier of speech recognition and translation,” https://nvidia.github.io/NeMo/blogs/ 2024/2024-02-canary/, 2024

    2024

  31. [40]

    Leveraging weakly supervised data to improve end-to-end speech-to-text translation,

    Ye Jia, Melvin Johnson, Wolfgang Macherey, Ron J Weiss, Yuan Cao, Chung-Cheng Chiu, Naveen Ari, Stella Laurenzo, and Yonghui Wu, “Leveraging weakly supervised data to improve end-to-end speech-to-text translation,” in ICASSP 2019-2019 IEEE International Conference on Acoustics...

    2019

  32. [41]

    Is chatgpt a good translator? yes with gpt-4 as the engine,

    Wenxiang Jiao, Wenxuan Wang, Jen-tse Huang, Xing Wang, Shuming Shi, and Zhaopeng Tu, “Is chatgpt a good translator? yes with gpt-4 as the engine,” arXiv preprint arXiv:2301.08745, 2023

    2023 arXiv

  33. [42]

    Document-level machine translation with large lan- guage models,

    Longyue Wang, Chenyang Lyu, Tianbo Ji, Zhirui Zhang, Dian Yu, Shuming Shi, and Zhaopeng Tu, “Document-level machine translation with large lan- guage models,” arXiv preprint arXiv:2304.02210, 2023

    2023 arXiv

  34. [43]

    Bleu: a method for automatic evaluation of machine translation,

    Kishore Papineni, Salim Roukos, Todd Ward, and Wei- Jing Zhu, “Bleu: a method for automatic evaluation of machine translation,” in Proceedings of the 40th Annual Meeting of the Association for Computational Linguis- tics, Pierre Isabelle, Eugene Charniak, and Dekang Lin, Eds.,...

    2002

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.