Pith. sign in

REVIEW 4 major objections 5 minor 38 references

Multilingual Gloss-free Sign Language Translation: Towards Building a Sign Language Foundation Model

T0 review · 4 major / 5 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 A single gloss-free model with dual CTC objectives translates ten sign languages across one-to-one, many-to-one, and many-to-many settings, while resolving language conflicts.

desk verdict First gloss-free multilingual SLT benchmark, but the LIDtok novelty is a word-count prior, not an alignment. read the letter →

arxiv 2505.24355 v1 pith:ZVGJL57R submitted 2025-05-30 cs.CL

Sihan Tan , Taro Miyazaki , Kazuhiro Nakadai This is my paper
classification cs.CL
keywords signlanguagetranslationmultilingualgloss-freeconnectionisttemporalclassificationtoken-levelidentificationmany-to-manySP-10PHOENIX14T
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

This paper aims to establish that one multilingual, gloss-free model can translate sign language videos from ten sign languages into their paired spoken languages under three different task configurations: one-to-one, many-to-one, and many-to-many. The proposed architecture, Sign2(LID+Text), adds two auxiliary CTC objectives to a Transformer encoder: a token-level sign language identification loss (LIDtok) on the initial encoder layer and a text CTC loss (TxtCTC) on the final encoder layer. The paper reports that this design mitigates the language conflicts that otherwise cause a 1.50 BLEU drop in a universal many-to-one baseline, and that it achieves competitive or leading results on SP-10, PHOENIX14T, and CSL-Daily. A sympathetic reader would care because gloss-free and multilingual are the two properties needed to make sign language translation practical at scale: no gloss annotations are required, and one model can serve many language pairs.

What carries the argument

The load-bearing mechanism is a hierarchical encoder supervised by two auxiliary CTC (Connectionist Temporal Classification) objectives. The initial encoder layer is supervised by a token-level sign language identification loss ($L_{\mathrm{LID}}$) that predicts a sequence of language IDs whose length equals the number of words in the target spoken text. The final encoder layer is supervised by a text CTC loss ($L_{\mathrm{Txt}}$) that aligns the latent sign representation with the spoken text, allowing the encoder to reorder non-monotonically while the attention decoder generates the translation with CTC as an auxiliary beam-search guide. The full objective is $L_{\mathrm{total}} = \lambda_1 L_{\mathrm{LID}} + \lambda_2 L_{\mathrm{Txt}} + \lambda_3 L_{\mathrm{Attn}}$ with weights $1$, $5$, and $3$ in the experiments.

What would settle it

A concrete falsifier would be to shuffle the token-level sign-language ID labels within each training utterance and retrain the many-to-one model: if test BLEU does not fall measurably, then LIDtok is not the mechanism resolving language conflict. A more direct check is to inspect the initial encoder's CTC alignment posteriors and test whether language-ID tokens switch at points matching visible sign boundaries; if the alignments are arbitrary, the auxiliary supervision is not doing the alignment work the paper attributes to it.

Watch

Extended reading notes

Core claim

The central claim is that token-level sign language identification is the missing mechanism for multilingual gloss-free translation. Instead of one utterance-level language label, the model predicts a sequence of language IDs with the same length as the target spoken text, so every word position receives a fine-grained language cue; a second text-oriented CTC objective then reorders the latent sign representations into the spoken sentence within a joint CTC/attention decoder. On the paper's own numbers, the resulting model beats the existing multilingual baseline on many-to-one SP-10 translation (5.18 test BLEU versus 4.66 for MLSLT), keeps many-to-many quality close to one-to-one quality up to five language pairs, and improves one-to-one test BLEU over its vanilla baseline on PHOENIX14T (24.23 versus 22.52) and CSL-Daily (14.18 versus 11.76). The paper also presents the first gloss-free one-to-one SLT results on SP-10 for all ten languages, thereby giving the field a common reference point for multilingual gloss-free work.

Load-bearing premise

The method assumes that the model can learn a reliable CTC alignment between the first encoder layer's output and a token-level sequence of sign-language IDs whose length equals the number of words in the spoken text, even though sign languages have no explicit word boundaries and may order words differently from the spoken language.

Editorial extensions

If this is right

  • If the dual-CTC design is correct, one model can serve ten sign languages without gloss supervision, removing the need to annotate glosses for every new language.
  • The many-to-many results imply that cross-lingual transfer can help low-resource sign languages: translation quality stays close to one-to-one quality when up to five language pairs are trained together.
  • Token-level language identification is more helpful as the number of mixed languages grows, so the benefit should increase as the model scales toward a true sign language foundation model.
  • The paper's one-to-one SP-10 benchmark, the first gloss-free multilingual benchmark on that corpus, gives later work a fixed point of comparison across all ten languages.
  • Since TxtCTC helps most on short and medium-length sentences, the alignment gains should concentrate where frame-to-token correspondences are clearest rather than on long, globally-structured sentences.

Reading between the lines

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

  • The token-level LID supervision may actually be learning sign subunits or articulatory segments rather than word-sized units; if so, the same objective could double as an unsupervised segmentation signal for sign language recognition.
  • Because the LID sequence length is tied to the spoken text length, the method could fail for sign languages whose word order diverges sharply from the spoken language; testing it on verb-final versus verb-initial sign language pairs would reveal where the monotonicity prior breaks.
  • The dual-CTC recipe mirrors successful multilingual speech recognition designs, so a plausible next step is to replace the hand-assigned LID labels with continuous language embeddings and train the text-CTC head on far larger multilingual corpora, which the paper does not itself attempt.
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 proposes Sign2(LID+Text), a multilingual gloss-free sign language translation model with a hierarchical encoder trained with two auxiliary CTC objectives: a token-level sign language identification (LIDtok) objective applied at an early encoder layer and a text CTC objective applied at the final encoder layer. The model is evaluated in one-to-one, many-to-one, and many-to-many settings on SP-10, PHOENIX14T, and CSL-Daily, and the authors report competitive BLEU/ROUGE scores, including a many-to-one improvement over a universal baseline and a new one-to-one benchmark on SP-10. The paper also releases code and model weights.

Significance. If the empirical claims held, the paper would be a useful step toward a multilingual gloss-free SLT model, and the released code plus the first one-to-one SP-10 benchmark would be valuable community resources. The many-to-one comparison against MLSLT and the incremental many-to-many experiments address an underexplored problem. However, the central novelty—token-level sign language identification—is not well supported by the construction in Eq. (2), and several empirical claims are overstated relative to the tables, so the significance depends on substantial revision.

major comments (4)
  1. [§3.2, Eq. (2)] The LIDtok target Itok is defined as the same language ID repeated |T| times, where |T| is the length of the spoken text. Under CTC, repeated identical labels must be separated by blanks, so this objective can only supervise the number of output positions, not align specific sign representations to specific spoken words. The statement in §3.2 that the |T|-length LIDtok sequence 'explicitly aligns the sign representations with each spoken word' is therefore not supported by the construction. The paper should either provide evidence that the CTC path emits LID tokens at meaningful positions (e.g., analysis of intermediate predictions or attention maps) or reframe the contribution as a length-prior/word-count objective rather than token-level identification.
  2. [§5, many-to-one, Table 4] The many-to-one section attributes the improvement over the universal baseline to mitigation of language conflicts, but Table 4 compares the full Sign2(LID+Text) model only against an individual/universal baseline and prior methods. There is no ablation that removes LIDtok while keeping TxtCTC, even though one-to-one experiments already show that TxtCTC alone yields sizable gains. Without this ablation, the improvement cannot be attributed to Sign2LID, and the conclusion in §6 that 'Sign2LID effectively mitigates language conflicts' is unsupported.
  3. [Appendix E, Table 9] The ablation study in Table 9 shows that LIDtok hurts performance at the 2-pair and 3-pair settings (7.48 vs. 6.22 and 6.50 vs. 6.15 BLEU, respectively) before helping at 4 pairs and beyond. Since the paper reports no error bars or significance tests on a very small dataset (830 training samples per language), these differences may be within noise. The claim that 'LIDtok is especially effective under more challenging translation conditions' requires a statistical basis and an explanation of the harmful regime at small pair counts.
  4. [§1, contribution statement vs. Table 3] The introduction claims the model 'achieve[s] state-of-the-art performance for each task,' but Table 3 shows that on CSL-Daily the proposed model (BLEU 14.18) is below SignLLM (BLEU 15.75), and similar discrepancies exist in ROUGE. The abstract more cautiously says 'competitive performance,' but the contribution statement is not consistent with the reported results. The claim should be revised to match the tables, or the comparison should be explained (e.g., different training data or decoding settings).
minor comments (5)
  1. [Table 4] In the Dev-ROUGE block, the mean value for 'Individual (10)' is reported as 28.53, but the individual ROUGE values listed range from roughly 27.70 to 36.60, whose arithmetic mean is about 32.16. This appears to be a typographical error and should be corrected.
  2. [Table 5] The text says the many-to-many model 'maintains comparable performance as the number of language pairs increases to five,' but the BLEU gap between one-to-one and many-to-many grows from 0.06 at 2 pairs to 0.97 at 5 pairs. The wording should be softened or the relevant pairwise significances reported.
  3. [Appendix E / Table 9] Please report whether the reported BLEU scores are averaged over multiple random seeds; for a dataset as small as SP-10, single-run results are not sufficient to judge the direction of the ablation.
  4. [Notation] The model name is inconsistently rendered as 'Sign2(LID+Text)' in the abstract and main text but 'Sign2LID' in §3.2 and 'Sign2 (LID+Text)' in Table 1. Please standardize the notation.
  5. [References] The paper cites 'Higuchi et al. (2022)' for hierarchical conditioning and 'Yan et al. (2023)' for joint CTC/attention decoding; both are appropriate, but the text would benefit from a brief sentence clarifying how these prior works were adapted to the sign-language setting, particularly the choice of the initial encoder layer for LIDtok.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central multilingual SLT results are empirical against external held-out benchmarks, and neither the LIDtok auxiliary loss nor the single self-citation makes a reported quantity equal to a fitted input by construction.

full rationale

The paper's claims are validated with held-out test partitions of three external benchmarks (SP-10, PHOENIX14T, CSL-Daily) using BLEU/ROUGE, so the reported numbers are not constructed from the model's inputs. The only candidate circular-construction concern is the token-level LID objective: Eq. 2 defines Itok as the same sign-language ID repeated |T| times, where |T| is the length of the target spoken sentence. That means LIDtok supervision carries language identity plus a word-count prior, not evidence of true word-level sign alignment, and the claim in Section 3.2 that it 'explicitly aligns the sign representations with each spoken word' overstates what Eq. 2 provides. This is a modeling/validity critique of the auxiliary supervision, not circularity: the loss is a training regularizer, and the measured effect (Table 9, many-to-one results) is empirically determined on dev/test, not equal by construction to a fitted parameter. The self-citation to Tan et al. (2025) is used only as provenance for the joint CTC/Attention architecture and text-CTC objective; it is prior published work independently re-evaluated here against external baselines, and it is not invoked as a uniqueness theorem or as the justification for the multilingual claim. No renamed known result is presented as organization. The Limitations section openly acknowledges data scarcity and small vocabulary, which are external constraints rather than circular steps.

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

The method introduces no new physical or semantic entities. Its commitments are modeling choices (loss weights, layer assignment, pair ordering) and domain assumptions about the pretrained feature extractor and CTC alignment, all stemming from Section 3 and Appendix B.

free parameters (3)
  • Loss weights lambda1, lambda2, lambda3 = 1, 5, 3
    Set in Appendix B without sensitivity analysis; the balance between LID CTC, text CTC, and attention loss is chosen by hand and directly affects the training objective.
  • Encoder layer assigned to Sign2LID = initial (first) layer
    Section 3.2 assigns the auxiliary LID task to the initial encoder layer based on the hierarchical conditioning intuition; no experiment compares different layer assignments.
  • Ordering of language pairs in many-to-many = by descending dev BLEU (Appendix F)
    The incremental scaling experiment adds pairs in order of one-to-one dev performance; this favorable ordering is a selection choice, and the full 10x10 matrix is not tested.
assumptions (4)
  • standard math CTC can align sequences of different lengths with monotonicity constraints.
    Used as the basis for both LIDtok and TxtCTC objectives (Section 3.2, Eq. 2 and Eq. 4).
  • domain assumption The pretrained SlowFastSign feature extractor provides transferable sign embeddings across all ten SP-10 languages.
    Section 3.1 states the extractor is 'pre-trained on spoken texts' but gives no details on pretraining data; the entire model relies on this shared visual encoder.
  • ad hoc to paper A token-level sequence of sign language IDs can be aligned to the spoken text words in the training labels.
    Table 1 defines the training label for token-level SLI as replacing all tokens with LIDs; this assumes each spoken word position has a meaningful SL ID, which is not linguistically justified.
  • domain assumption SP-10, PHOENIX14T, and CSL-Daily are representative benchmarks for evaluating multilingual SLT.
    The results and conclusions are drawn from these three datasets; SP-10 is noted as the only public multilingual SLT corpus (Limitations).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multilingual Gloss-free Sign Language Translation: Towards Building a Sign Language Foundation Model." pith.science (2026). https://pith.science/paper/ZVGJL57R

@misc{pith2026250524355,
  author       = {Pith},
  title        = {Pith review of: Multilingual Gloss-free Sign Language Translation: Towards Building a Sign Language Foundation Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZVGJL57R}},
  note         = {Machine review of arXiv:2505.24355}
}
read the original abstract

Sign Language Translation (SLT) aims to convert sign language (SL) videos into spoken language text, thereby bridging the communication gap between the sign and the spoken community. While most existing works focus on translating a single sign language into a single spoken language (one-to-one SLT), leveraging multilingual resources could mitigate low-resource issues and enhance accessibility. However, multilingual SLT (MLSLT) remains unexplored due to language conflicts and alignment difficulties across SLs and spoken languages. To address these challenges, we propose a multilingual gloss-free model with dual CTC objectives for token-level SL identification and spoken text generation. Our model supports 10 SLs and handles one-to-one, many-to-one, and many-to-many SLT tasks, achieving competitive performance compared to state-of-the-art methods on three widely adopted benchmarks: multilingual SP-10, PHOENIX14T, and CSL-Daily.

Figures

Figures reproduced from arXiv: 2505.24355 by the authors.

Figure 1
Figure 1. Overview of multilingual gloss-free model. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of multilingual gloss-free model. [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Average BLEU score on different token length [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Average BLEU score on different token length [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Sign language similarities and differences [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

38 extracted references · 21 canonical work pages

  1. [1]

    Junseok Ahn, Youngjoon Jang, and Joon Son Chung. 2024. https://doi.org/10.1109/ICASSP48485.2024.10445841 Slowfast network for continuous sign language recognition . In ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 3920--3924

    arXiv 2024

  2. [2]

    Necati Cihan Camgoz, Simon Hadfield, Oscar Koller, Hermann Ney, and Richard Bowden. 2018. Neural sign language translation. In Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR), pages 7784--7793

    work page 2018

  3. [3]

    Necati Cihan Camgoz, Oscar Koller, Simon Hadfield, and Richard Bowden. 2020. Sign Language Transformers : Joint end-to-end sign language recognition and translation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 10023--10033

    work page 2020

  4. [4]

    William Chen, Brian Yan, Jiatong Shi, Yifan Peng, Soumi Maiti, and Shinji Watanabe. 2023. Improving massively multilingual asr with auxiliary ctc objectives. In ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1--5. IEEE

    work page 2023

  5. [5]

    Yutong Chen, Ronglai Zuo, Fangyun Wei, Yu Wu, Shujie LIU, and Brian Mak. 2022. https://proceedings.neurips.cc/paper_files/paper/2022/file/6cd3ac24cdb789beeaa9f7145670fcae-Paper-Conference.pdf Two-stream network for sign language recognition and translation . In Advances in Neural Information Processing Systems, volume 35, pages 17043--17056. Curran Associ...

    work page 2022

  6. [6]

    Zhigang Chen, Benjia Zhou, Jun Li, Jun Wan, Zhen Lei, Ning Jiang, Quan Lu, and Guoqing Zhao. 2024. https://aclanthology.org/2024.lrec-main.620/ Factorized learning assisted with large language model for gloss-free sign language translation . In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Eval...

    work page 2024

  7. [7]

    Binyam Gebrekidan Gebre, Peter Wittenburg, and Tom Heskes. 2013. https://doi.org/10.1109/ICIP.2013.6738541 Automatic sign language identification . In 2013 IEEE International Conference on Image Processing, pages 2626--2630

    work page arXiv 2013

  8. [8]

    Jia Gong, Lin Geng Foo, Yixuan He, Hossein Rahmani, and Jun Liu. 2024. Llms are good sign language translators. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 18362--18372

    2024

Show all 38 references
  1. [9]

    Alex Graves, Santiago Fern\' a ndez, Faustino Gomez, and J\" u rgen Schmidhuber. 2006. https://doi.org/10.1145/1143844.1143891 Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks . In International Conference on Machine Lea...

    2006

  2. [10]

    Shester Gueuwou, Sophie Siake, Colin Leong, and Mathias M \"u ller. 2023. https://doi.org/10.18653/v1/2023.findings-emnlp.664 JWS ign: A highly multilingual corpus of B ible translations for more diversity in sign language processing . In Findings of the Association for Comput...

    2023 doi

  3. [11]

    Yasser Hamidullah, Josef van Genabith, and Cristina Espa \ n a-Bonet. 2024. https://doi.org/10.18653/v1/2024.acl-short.40 Sign language translation with sentence embedding supervision . In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics ...

    2024 doi

  4. [12]

    Yosuke Higuchi, Keita Karube, Tetsuji Ogawa, and Tetsunori Kobayashi. 2022. Hierarchical conditional end-to-end asr with ctc and multi-granular subword units. In ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 7797--7801. IEEE

    2022

  5. [13]

    spreadthesign

    Marlene Hilzensauer and Klaudia Krammer. 2015. A multilingual dictionary for sign languages:" spreadthesign". In ICERI2015 Proceedings, pages 7826--7834. IATED

    2015

  6. [14]

    Zifan Jiang, Gerard Sant, Amit Moryossef, Mathias M \"u ller, Rico Sennrich, and Sarah Ebling. 2024. https://doi.org/10.18653/v1/2024.emnlp-main.518 S ign CLIP : Connecting text and sign language by contrastive learning . In Proceedings of the 2024 Conference on Empirical Meth...

    2024 doi

  7. [15]

    Le, Maxim Krikun, Yonghui Wu, Zhifeng Chen, Nikhil Thorat, Fernanda Vi \'e gas, Martin Wattenberg, Greg Corrado, Macduff Hughes, and Jeffrey Dean

    Melvin Johnson, Mike Schuster, Quoc V. Le, Maxim Krikun, Yonghui Wu, Zhifeng Chen, Nikhil Thorat, Fernanda Vi \'e gas, Martin Wattenberg, Greg Corrado, Macduff Hughes, and Jeffrey Dean. 2017. https://doi.org/10.1162/tacl_a_00065 G oogle`s multilingual neural machine translatio...

    2017 doi

  8. [16]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. 2015. http://arxiv.org/abs/1412.6980 Adam: A method for stochastic optimization . In 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings

    2015 arXiv

  9. [17]

    Liddell and Robert E

    Scott K. Liddell and Robert E. Johnson. 1989. http://www.jstor.org/stable/26204052 American sign language: The phonological base . Sign Language Studies, (64):195--278

    1989

  10. [18]

    Chin-Yew Lin. 2004. ROUGE : A package for automatic evaluation of summaries. In Text summarization branches out, pages 74--81

    2004

  11. [19]

    Mathias M \"u ller, Zifan Jiang, Amit Moryossef, Annette Rios, and Sarah Ebling. 2023. https://doi.org/10.18653/v1/2023.acl-short.60 Considerations for meaningful sign language machine translation based on glosses . In Proceedings of the 61st Annual Meeting of the Association ...

    2023 doi

  12. [20]

    Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. BLEU : A method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting of the Association for Computational Linguistics, pages 311--318

    2002

  13. [21]

    Matt Post. 2018. https://doi.org/10.18653/v1/W18-6319 A call for clarity in reporting BLEU scores . In Proceedings of the Third Conference on Machine Translation: Research Papers, pages 186--191, Brussels, Belgium. Association for Computational Linguistics

    2018 doi

  14. [22]

    Ramon Sanabria and Florian Metze. 2018. https://doi.org/10.1109/SLT.2018.8639530 Hierarchical multitask learning with ctc . In 2018 IEEE Spoken Language Technology Workshop (SLT), pages 485--490

    2018

  15. [23]

    Sihan Tan, Taro Miyazaki, Nabeela Khan, and Kazuhiro Nakadai. 2025. https://aclanthology.org/2025.coling-main.219/ Improvement in sign language translation using text CTC alignment . In Proceedings of the 31st International Conference on Computational Linguistics, pages 3255--...

    2025

  16. [24]

    Garrett Tanzer. 2024. https://arxiv.org/abs/2408.13585 Fleurs-asl: Including american sign language in massively multilingual multitask evaluation . Preprint, arXiv:2408.13585

    2024 arXiv

  17. [25]

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, ukasz Kaiser, and Illia Polosukhin. 2017. https://proceedings.neurips.cc/paper_files/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf Attention is all you need . In Advances in Ne...

    2017

  18. [26]

    Fangyun Wei and Yutong Chen. 2023. Improving continuous sign language recognition with cross-lingual signs. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 23612--23621

    2023

  19. [27]

    Ryan Wong, Necati Cihan Camgoz, and Richard Bowden. 2024. https://openreview.net/forum?id=LqaEEs3UxU Sign2 GPT : Leveraging large language models for gloss-free sign language translation . In The Twelfth International Conference on Learning Representations

    2024

  20. [28]

    Brian Yan, Siddharth Dalmia, Yosuke Higuchi, Graham Neubig, Florian Metze, Alan W Black, and Shinji Watanabe. 2023. https://doi.org/10.18653/v1/2023.eacl-main.119 CTC alignments improve autoregressive translation . In Proceedings of the 17th Conference of the European Chapter ...

    2023 doi

  21. [29]

    Aoxiong Yin, Zhou Zhao, Weike Jin, Meng Zhang, Xingshan Zeng, and Xiaofei He. 2022. https://doi.org/10.1109/CVPR52688.2022.00505 Mlslt: Towards multilingual sign language translation . In 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 5099--5109

    2022

  22. [30]

    Kayo Yin and Jesse Read. 2020. https://doi.org/10.18653/v1/2020.coling-main.525 Better sign language translation with STMC -transformer . In Proceedings of the 28th International Conference on Computational Linguistics, pages 5975--5989, Barcelona, Spain (Online). Internationa...

    2020 doi

  23. [31]

    Biao Zhang, Barry Haddow, and Rico Sennrich. 2022. Revisiting end-to-end speech-to-text translation from scratch. In International Conference on Machine Learning, pages 26193--26205. PMLR

    2022

  24. [32]

    Biao Zhang, Mathias M \"u ller, and Rico Sennrich. 2023. https://openreview.net/forum?id=EBS4C77p_5S SLTUNET : A simple unified model for sign language translation . In The Eleventh International Conference on Learning Representations

    2023

  25. [33]

    Ruiquan Zhang, Cong Hu, Pei Yu, and Yidong Chen. 2025. https://aclanthology.org/2025.coling-main.241/ Improving multilingual sign language translation with automatically clustered language family information . In Proceedings of the 31st International Conference on Computationa...

    2025

  26. [34]

    Jian Zhao, Weizhen Qi, Wengang Zhou, Nan Duan, Ming Zhou, and Houqiang Li. 2022. https://doi.org/10.1109/TMM.2021.3087006 Conditional sentence generation and cross-modal reranking for sign language translation . IEEE Transactions on Multimedia, 24:2662--2672

    2022

  27. [35]

    Benjia Zhou, Zhigang Chen, Albert Clap\'es, Jun Wan, Yanyan Liang, Sergio Escalera, Zhen Lei, and Du Zhang. 2023. Gloss-free sign language translation: Improving from visual-language pretraining. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)...

    2023

  28. [36]

    Hao Zhou, Wengang Zhou, Weizhen Qi, Junfu Pu, and Houqiang Li. 2021. Improving sign language translation with monolingual data by sign back-translation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 1316--1325

    2021

  29. [37]

    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...

  30. [38]

    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.