Pith.

Computation and Language · arXiv:2611.16359 v1 · submitted 12 Nov 2026 · 72 references

IsharaBART: An efficient BART-Based Framework for Low-Resource Bangla Sign Language Translation

Aonmoy Das, Ananna Dev Aishi, Md. Fazlul Kader

The paper says

“Millions of Bangla Sign Language (BdSL) users lack reliable automatic translation tools. Existing systems are limited by data scarcity, unidirectional design, and computationally expensive architectures that are not suitable for low-resource deployment.”

more of the abstract

We introduce IsharaBART, a bidirectional generative framework for gloss-sentence translation in BdSL. IsharaBART uses Bidirectional and Auto-Regressive Transformers (BART)-based conditional language modeling with denoising learning objectives. We developed a parallel dataset of 1,000 professionally annotated gloss-sentence pairs, balanced across tenses and structures. We expanded the dataset to 4,831 pairs through four augmentation techniques: rule-based tense transformations, masked token substitutions, back-translation, and one-shot prompting. A compact denoising sequence-to-sequence model was pre-trained from scratch on monolingual Bangla text from six different corpora, then fine-tuned on the annotated and augmented datasets. We compare IsharaBART against four larger baseline models under identical conditions, where it achieves comparable or superior bidirectional performance. On the gloss-to-sentence (G2S) task, it attains Bilingual Evaluation Understudy (BLEU-4) score up to 51.7, and on the sentence-to-gloss (S2G) task, it achieves 61.40, outperforming multilingual models despite having fewer than 10% of the parameters. Ablation studies demonstrate that pre-training provides 62.5-90.4% faster convergence and 11.5-118% higher BLEU-4 than scratch. Low-rank adaptation (LoRA) reduces trainable parameters by 99.3%, but with 28.0-52.2% loss of BLEU-4 compared to full-parameter fine-tuning.

from the abstract

Pith says

Machine review · deepseek-v4-flash · 13 Nov 2026 · what was read was not recorded · not checked by a person

It can be wrong. Challenge this review or reply to it.

A useful small-model recipe for Bangla Sign Language gloss–sentence translation, with an evaluation-integrity question that needs answering before the headline numbers are trusted. This line is wrong

the referee pass did not run

How often objections from reviews like this one are right: not yet measured

Are you an author? What you can do.

Not checked: figures, code, what the 72 references say.

What would settle it

Obtain the exact train/validation/test splits. For each test pair, compute the maximum lexical overlap (BLEU-4 or 4-gram F1) with every training pair, and count test pairs that are near-duplicates of a training pair — i.e., derived from the same seed sentence via tense change, single-token substitution, or gloss-preserving paraphrase. Then re-fine-tune all five models under a seed-grouped split (all augmented derivatives of each original annotated sentence kept in one partition) with three random seeds, reporting mean ± SD BLEU-4. If the near-duplicate fraction is non-trivial, or if IsharaBART's S2G BLEU-4 no longer exceeds mBART-50's under the grouped split, the headline comparison is unsupported.

from the skeptic's reading

How the review went

3 readings by deepseek-v4-flash · 13 Nov 2026

  1. Reader holds up, with conditions confidence moderate
  2. Skeptic partly agrees 53 s · 32,663 tokens read, 8,098 written
  3. Editor wrote the verdict above 31 s · 33,639 tokens read, 3,992 written

Not run on this paper: the referee, the simulated authors' defence.

Pith did not check this review's objections against the paper.

The review in full

The review in brief, machine-written

Short version: this is a serious, mostly careful low-resource NLP paper, but the central claim about beating mBART-50 is thinner and more fragile than the authors make it sound. The evaluation split is never described, and that matters more than the small dataset.

What is new and good: they train a 56M monolingual BART from scratch on Bangla text, fine-tune bidirectionally on 1,000 professional gloss-sentence pairs plus 3,831 augmented pairs, and compare against four baselines under controlled conditions with ablations for pretraining and LoRA. The convergence and parameter-efficiency numbers are real evidence: pretraining cuts epochs dramatically, and full fine-tuning beats LoRA by large margins. The dataset itself is a contribution, albeit small. The paper is honest about data size, lack of video integration, and unmeasured deployment costs.

Soft spots, in order of size. First, the data split is unspecified. Section 4.4 just says results are on a held-out test set; there is no description of how the 4,831 pairs were split, no duplicate screening, no mention of whether augmented variants of the same seed sentence can span train and test. The augmentation methods are near-duplicate generators: tense transformation changes verb inflection, masked substitution swaps one word, back-translation keeps the original gloss, and the LLM prompt says to follow the same grammatical style. With 100 words and roughly ten sentence templates per word, a random split is likely to leak template variants into the test set. The headline S2G "outperformance" over mBART-50 rests on a 0.58 BLEU-4 margin; if any of that comes from template memorization, the claim disappears. This is not a data-quality judgment; it is an absent sentence in the methods that can be fixed. The authors should release the split, or better, the data and code.

Second, the human dataset evaluation scores are very high (99.21 adequacy, 97.7% both criteria). Two raters, no adjudication, moderate kappa. Not fatal, but it does not independently establish the dataset quality claim. Third, Table 12 compares BLEU across different datasets and languages; it is illustrative at best and should not be read as a direct comparison. Fourth, no code or data release is mentioned, which limits reproducibility. The lack of comparison to Dhrubo et al. [8] is minor since that system operates on video/landmarks, not gloss-sentence text. The authors already acknowledge the small dataset in Section 6, so I will not relitigate that.

Bottom line: the experiment design is mostly sound and the paper deserves serious peer review. But the evaluation split issue needs to be resolved before the S2G superiority claim can stand. I would ask for clarification of the split construction and duplication screening before accepting.

The strongest objection

The central claim — S2G 61.40 "outperforming" mBART-50's 60.82 — is a 0.58-point BLEU-4 margin whose validity requires the held-out test set to be genuinely disjoint from training content. Section 4.4 states results are on a held-out test set, but the paper never describes how the 4,831 pairs were split. This matters because the augmentation methods in §3.2 are explicitly near-duplicate generators: rule-based tense transformation changes only the verb inflection, masked token substitution swaps a single content word, back-translation paraphrases one sentence while keeping the original gloss, and the one-shot prompt (Table 2) instructs the LLM to "follow the same grammatical style." With only 100 words arranged in roughly ten sentences per word, a random pair-level split will place variants of the same seed sentence in both train and test. The model could then score highly by memorizing the 1,000 seed templates rather than by generalization, inflating the absolute numbers (51.7 G2S, 61.4 S2G) and easily flipping the sub-1-point S2G margin over mBART-50 or the 0.50-point G2S gap to BanglaT5. The Table 12 comparison to PHOENIX-14T is likewise only meaningful if the test set is not templated from training seeds. This is an evaluation-integrity question, not a data-quality judgment, and the information needed to resolve it (split construction, duplication screening) is absent from the paper.

The reader's CONDITIONAL verdict is the right one, and this pass sharpens rather than overturns it. The paper has genuine independent support: a controlled 5×2×2 fine-tuning design, a pre-training ablation with convergence curves, honest reporting of the G2S gap to BanglaT5 (51.74 vs 52.24), and a human evaluation whose own statistics (two evaluators, kappa 0.364–0.495) temper its headline scores. Section 6 states the main limitations explicitly. The load-bearing condition for the strongest claim is not data quality per se but evaluation integrity: the test set must be free of near-duplicate augmented variants of training seeds. The paper does not describe the split construction (§4.3–4.4 cover conditions and metrics only), and every augmentation method in §3.2 is engineered to produce pairs closely tied to a seed sentence (tense inflection, one-token swap, gloss-preserving round-trip paraphrase, template-following LLM generation). With 100 words and roughly ten sentences per word, even a random split yields train/test near-duplicates. The headline margins (0.58 S2G over mBART-50, 0.50 G2S below BanglaT5) are smaller than typical run-to-run variance, and no seeds or significance tests are reported for the baseline comparisons. The concrete test (duplication screening plus a seed-grouped, multi-seed re-run) settles the concern without requiring re-derivation of any external result. If it fails, the numbers reflect split construction, not model capability; if it passes, the claim stands. This does not change the CONDITIONAL verdict; it makes the first condition specific and checkable.

Pith's reading of the paper (a machine's summary, not the authors')

The paper introduces IsharaBART, a small (56M parameter) BART-based model for bidirectional translation between Bangla Sign Language (BdSL) gloss sequences and written Bangla sentences. The authors aim to show that a compact, monolingually pre-trained model can match or exceed the performance of much larger multilingual models (e.g., mBART-50 with 611M parameters) on this low-resource task, provided it is pre-trained on diverse Bangla text and fine-tuned on a carefully augmented dataset. They construct a parallel gloss-sentence dataset of 4,831 pairs (1,000 professionally annotated plus 3,831 synthetically augmented) and demonstrate that pre-training yields 62.5-90.4% faster convergence and 11.5-118% higher BLEU-4 compared to training from scratch. If true, this suggests that targeted monolingual pre-training and data augmentation can enable effective sign language translation even with very limited parallel data and computational resources, offering a practical path for under-resourced sign languages.

What carries the argument. The central object is a BART (Bidirectional and Auto-Regressive Transformer) encoder-decoder model with 4 layers each, 512 hidden dimensions, 8 attention heads, and approximately 56M parameters. The key mechanism is denoising pre-training on 60 GB of monolingual Bangla text from six corpora, using a composite corruption function that applies sentence permutation, text infilling (Poisson span masking), token deletion, token masking, and word permutation. This pre-training teaches the model to reconstruct original text from various corrupted forms, which the authors argue is particularly beneficial for learning robust mappings between the two modalities (glosses and sentences) in sign language translation. The model is then fine-tuned on the parallel gloss-sentence dataset for both translation directions.

Core claim. A 56M-parameter BART model pre-trained from scratch on monolingual Bangla text and fine-tuned on a small parallel gloss-sentence dataset (4,831 pairs) achieves BLEU-4 scores of 51.7 for gloss-to-sentence (G2S) and 61.4 for sentence-to-gloss (S2G) translation in Bangla Sign Language. This outperforms the much larger mBART-50 (611M parameters) on S2G and comes within 0.5 BLEU-4 of BanglaT5 (247M parameters) on G2S, while using 77-91% fewer parameters. The authors demonstrate that monolingual pre-training with denoising objectives provides substantial gains in both translation quality and convergence speed, and that four complementary data augmentation techniques (rule-based tense transformation, masked token substitution, back-translation, and one-shot LLM prompting) effectively expand the limited parallel data while preserving semantic and structural consistency.

The premise it rests on. The 1,000 professionally annotated gloss-sentence pairs and the 3,831 synthetically augmented pairs are representative enough of real BdSL usage to train a model that generalizes beyond the limited vocabulary (100 words) and syntactic structures covered.

If the paper is right (Pith's inferences, not the authors')

Compact (56M parameter) models can achieve competitive sign language translation performance, making deployment on edge devices and mobile platforms feasible.

Monolingual pre-training on general-domain text provides substantial benefits for low-resource gloss-sentence translation, reducing the need for large parallel datasets.

Data augmentation techniques (tense transformation, masked substitution, back-translation, one-shot prompting) can effectively expand limited parallel data for sign language translation while preserving quality.

The bidirectional evaluation reveals asymmetric difficulty: sentence-to-gloss (S2G) benefits more from augmentation (22.0 BLEU-4 gain) than gloss-to-sentence (G2S) (8.4 BLEU-4 gain), suggesting different underlying challenges.

LoRA fine-tuning severely degrades performance (28-52% BLEU-4 loss) compared to full fine-tuning, indicating that parameter-efficient adaptation may be insufficient for specialized sequence-to-sequence tasks in low-resource settings.

The success of monolingual pre-training suggests that for low-resource sign languages, investing in monolingual text corpora and training small models from scratch may be more cost-effective than adapting large multilingual models.

The large asymmetry between G2S and S2G gains from augmentation (22 vs 8 BLEU-4) implies that the bottleneck for S2G is learning the highly constrained gloss vocabulary and order, which augmentation directly addresses by providing more varied input-output mappings.

The 97.7% human evaluation satisfaction rate on the dataset, combined with only 66% (G2S) and 56% (S2G) of model outputs meeting both adequacy and fluency criteria, suggests a substantial gap between data quality and model generation quality that future work should target.

The LoRA results (99.3% parameter reduction but 28-52% BLEU-4 loss) challenge the assumption that low-rank adaptation preserves translation quality for specialized tasks, implying that full fine-tuning may be necessary even for small models in low-resource domains.

Check this review

Pith stored no record of this review's passes and took no fingerprint of it. Reviews made by code from 13 Nov 2026 on carry both.

How often reviews like this one are right

Reviews like this one: written by deepseek-v4-flash, where what was read was not recorded.

How often objections from reviews like this one are right: not yet measured

Every value above, in the review's public record (JSON)

Reviews by people

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review

How Pith checks papers · The full audit of this paper

The paper

Abstract

Millions of Bangla Sign Language (BdSL) users lack reliable automatic translation tools. Existing systems are limited by data scarcity, unidirectional design, and computationally expensive architectures that are not suitable for low-resource deployment. We introduce IsharaBART, a bidirectional generative framework for gloss-sentence translation in BdSL. IsharaBART uses Bidirectional and Auto-Regressive Transformers (BART)-based conditional language modeling with denoising learning objectives. We developed a parallel dataset of 1,000 professionally annotated gloss-sentence pairs, balanced across tenses and structures. We expanded the dataset to 4,831 pairs through four augmentation techniques: rule-based tense transformations, masked token substitutions, back-translation, and one-shot prompting. A compact denoising sequence-to-sequence model was pre-trained from scratch on monolingual Bangla text from six different corpora, then fine-tuned on the annotated and augmented datasets. We compare IsharaBART against four larger baseline models under identical conditions, where it achieves comparable or superior bidirectional performance. On the gloss-to-sentence (G2S) task, it attains Bilingual Evaluation Understudy (BLEU-4) score up to 51.7, and on the sentence-to-gloss (S2G) task, it achieves 61.40, outperforming multilingual models despite having fewer than 10% of the parameters. Ablation studies demonstrate that pre-training provides 62.5-90.4% faster convergence and 11.5-118% higher BLEU-4 than scratch. Low-rank adaptation (LoRA) reduces trainable parameters by 99.3%, but with 28.0-52.2% loss of BLEU-4 compared to full-parameter fine-tuning.

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.

abstract compiled · 2026-11-13

The authors' main claim, as the classifier picked it out 0.77

“On the gloss-to-sentence (G2S) task, it attains Bilingual Evaluation Understudy (BLEU-4) score up to 51.7, and on the sentence-to-gloss (S2G) task, it achieves 61.40, outperforming multilingual models despite having fewer than 10% of the parameters.”

How the main claim is stated: asserted 1.00

What the abstract itself states, as probabilities: claims first 0.07 · states limitation 0.63 · human participants 0.42 · controlled comparison 0.95 · randomized assignment 0.02 · states quantitative result 0.99 · compares with prior results 0.98 · mentions code or data availability 0.03

The abstract, sentence by sentence

This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.

SentenceRoleThe authors' own workStates the main conclusionStrength
Millions of Bangla Sign Language (BdSL) users lack reliable automatic translation tools. background 0.93 0.05 0.02 asserted 1.00
Existing systems are limited by data scarcity, unidirectional design, and computationally expensive architectures that are not suitable for low-resource deployment. gap 0.59 0.04 0.02 asserted 1.00
We introduce IsharaBART, a bidirectional generative framework for gloss-sentence translation in BdSL. contribution 1.00 0.89 0.11 asserted 1.00
IsharaBART uses Bidirectional and Auto-Regressive Transformers (BART)-based conditional language modeling with denoising learning objectives. method 0.98 0.71 0.04 asserted 1.00
We developed a parallel dataset of 1,000 professionally annotated gloss-sentence pairs, balanced across tenses and structures. method 0.83 0.97 0.03 asserted 1.00
We expanded the dataset to 4,831 pairs through four augmentation techniques: rule-based tense transformations, masked token substitutions, back-translation, and one-shot prompting. method 0.95 0.97 0.03 asserted 1.00
A compact denoising sequence-to-sequence model was pre-trained from scratch on monolingual Bangla text from six different corpora, then fine-tuned on the annotated and augmented datasets. method 1.00 0.95 0.04 asserted 0.99
We compare IsharaBART against four larger baseline models under identical conditions, where it achieves comparable or superior bidirectional performance. result 0.94 0.94 0.25 asserted 1.00
On the gloss-to-sentence (G2S) task, it attains Bilingual Evaluation Understudy (BLEU-4) score up to 51.7, and on the sentence-to-gloss (S2G) task, it achieves 61.40, outperforming multilingual models despite having fewer than 10% of the parameters. result 1.00 0.96 0.54 asserted 1.00
Ablation studies demonstrate that pre-training provides 62.5-90.4% faster convergence and 11.5-118% higher BLEU-4 than scratch. result 1.00 0.95 0.09 asserted 1.00
Low-rank adaptation (LoRA) reduces trainable parameters by 99.3%, but with 28.0-52.2% loss of BLEU-4 compared to full-parameter fine-tuning. result 0.98 0.94 0.06 asserted 1.00

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.94

What was examined, and what was not

Sources read

  • arXiv listing abstract v1

Not examined

  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers

Levels run

  • P abstract

Not run

  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

How the labels were made

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-13: 1 request, 8,485 input tokens.

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON

Figures (6)
Figure 1 of the paper
Figure 1. Proposed methodology for developing IsharaBART.
Figure 2 of the paper
Figure 2. Dataset validation using (a) sentence length distributions, (b) gloss-to-text length ratios, (c) UMAP
Figure 3 of the paper
Figure 3. Proposed framework for the pre-training with denoising objectives.
Figure 4 of the paper
Figure 4. IsharaBART pre-training loss convergence over 30k steps. Training loss (blue) and validation loss
Figure 5 of the paper
Figure 5. BLEU-4 convergence curves comparing pre-trained IsharaBART against from-scratch baselines
Figure 6 of the paper
Figure 6. Multi-dimensional performance comparison between pre-trained IsharaBART and from-scratch

Figures are the authors', reproduced from the paper.

References (72)
  1. [8] T. H. Dhrubo, M. T. Reza, S. Sahariar, W. A. Mukto, Transformer based sign-to- text translation for bangladeshi sign language, Scientific Reports 16 (1) (2026) 1154. doi:10.1038/s41598-025-30856-y on Pithdoi:10.1038/s41598-025-30856-y
  2. [1] B. O. Olusanya, A. C. Davis, H. J. Hoffman, Hearing loss grades and the international classification of functioning, disability and health, Bulletin of the World Health Orga- nization 97 (10) (2019) 725––728.doi:10.2471/blt.19.230367 on Pithdoi:10.2471/blt.19.230367
  3. [2] J. Guo, P. Li, T. Cohn, Bridging sign and spoken languages: Pseudo gloss generation for sign language translation, in: Advances in Neural Information Processing Systems 38, Neural Information Processing Systems Foundation, Inc. (NeurIPS), San Diego, California, USA and Mexico ... on Pith
  4. [3] M. De Coster, D. Shterionov, M. Van Herreweghe, J. Dambre, Machine translation from signed to spoken languages: state of the art and challenges, Universal Access in theInformationSociety23(3)(2024)1305––1331. doi:10.1007/s10209-023-00992-1 on Pithdoi:10.1007/s10209-023-00992-1
  5. [4] S. Tan, N. Khan, Z. An, Y. Ando, R. Kawakami, K. Nakadai, A review of deep learning- based approaches to sign language processing, Advanced Robotics 38 (23) (2024) 1649– –1667. doi:10.1080/01691864.2024.2442721. 26 on Pithdoi:10.1080/01691864.2024.2442721
  6. [5] A. Othman, M. Jemni, English-ASL gloss parallel corpus 2012: ASLG-PC12, in: Pro- ceedings of the LREC2012 5th Workshop on the Representation and Processing of Sign Languages: Interactions between Corpus and Lexicon, European Language Resources Association (ELRA), Istanbul, Tur... on Pith
  7. [6] H. Zhou, W. Zhou, W. Qi, J. Pu, H. Li, Improving sign language translation with monolingual data by sign back-translation, in: 2021 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), IEEE, Nashville, TN, USA, 2021, pp. 1316––1325. doi:10.1109/CVPR46437.2021.00137 on Pithdoi:10.1109/cvpr46437.2021.00137
  8. [7] N. C. Camgoz, S. Hadfield, O. Koller, H. Ney, R. Bowden, Neural sign language trans- lation, in: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, IEEE, Salt Lake City, UT, 2018, pp. 7784––7793.doi:10.1109/CVPR.2018.00812 on Pithdoi:10.1109/cvpr.2018.00812
  9. [9] M. Hossen, A. Govindaiah, S. Sultana, A. Bhuiyan, Bengali sign language recognition using deep convolutional neural network, in: 2018 Joint 7th International Conference on Informatics, Electronics & Vision (ICIEV) and 2018 2nd International Conference on Imaging, Vision & Patt... on Pithdoi:10.1109/iciev.2018.8640962
  10. [10] M. S. Islam, S. S. S. Mousumi, N. A. Jessan, A. S. A. Rabby, S. Abujar, S. A. Hos- sain, Ishara-bochon: The first multipurpose open access dataset for bangla sign lan- guage isolated digits, in: Recent Trends in Image Processing and Pattern Recogni- tion, Vol. 1035, Springer S... on Pith
  11. [11] S. Siddique, S. Islam, E. E. Neon, T. Sabbir, I. T. Naheen, R. Khan, Deep learning- based bangla sign language detection with an edge device, Intelligent Systems with Applications 18 (2023) 200224.doi:10.1016/j.iswa.2023.200224 on Pithdoi:10.1016/j.iswa.2023.200224
  12. [12] M. J. Raihan, M. I. Labib, A. A. J. Jim, J. J. Tiang, U. Biswas, A.-A. Nahid, Bengali- sign: A machine learning-based bengali sign language interpretation for deaf and non- verbal people, Sensors 24 (16) (2024) 5351.doi:10.3390/s24165351 on Pithdoi:10.3390/s24165351
  13. [13] A. Das, A. D. Aishi, M. U. Toha, M. F. Kader, Bangla sign language translator for deaf and speech impaired people using deep lstm, International Journal of Speech Technology 28 (2) (2025) 351––368.doi:10.1007/s10772-025-10184-2 on Pithdoi:10.1007/s10772-025-10184-2
  14. [14] Y. Zhang, Y. Han, Z. Zhu, X. Jiang, Y. Zhang, Artificial intelligence in sign language recognition: A comprehensive bibliometric and visual analysis, Computers and Electri- cal Engineering 120 (2024) 109854.doi:10.1016/j.compeleceng.2024.109854. 27 on Pithdoi:10.1016/j.compeleceng.2024.109854
  15. [15] B. Kang, S. Tripathi, T. Q. Nguyen, Real-time sign language fingerspelling recognition using convolutional neural networks from depth map, in: 2015 3rd IAPR Asian Con- ference on Pattern Recognition (ACPR), IEEE, Kuala Lumpur, Malaysia, 2015, pp. 136––140. doi:10.1109/ACPR.201... on Pithdoi:10.1109/acpr.2015.7486481
  16. [16] M. Oberweger, P. Wohlhart, V. Lepetit, Training a feedback loop for hand pose esti- mation, in: 2015 IEEE International Conference on Computer Vision (ICCV), IEEE, Santiago, Chile, 2015, pp. 3316––3324.doi:10.1109/ICCV.2015.379 on Pithdoi:10.1109/iccv.2015.379
  17. [17] J. Huang, W. Zhou, H. Li, W. Li, Sign language recognition using 3d convolutional neural networks, in: 2015 IEEE International Conference on Multimedia and Expo (ICME), IEEE, Turin, Italy, 2015, pp. 1––6.doi:10.1109/ICME.2015.7177428 on Pithdoi:10.1109/icme.2015.7177428
  18. [18] R. Rastgoo, K. Kiani, S. Escalera, Multi-modal deep hand sign language recognition in still images using restricted boltzmann machine, Entropy 20 (11) (2018) 809.doi: 10.3390/e20110809 on Pithdoi:10.3390/e20110809
  19. [19] C. Wei, W. Zhou, J. Pu, H. Li, Deep grammatical multi-classifier for continuous sign language recognition, in: 2019 IEEE Fifth International Conference on Multimedia Big Data (BigMM), IEEE, Singapore, Singapore, 2019, pp. 435––442.doi:10.1109/BigMM. 2019.00027 on Pithdoi:10.1109/bigmm
  20. [20] J. Hasanov, N. Alishzade, A. Nazimzade, S. Dadashzade, T. Tahirov, Development of a hybrid word recognition system and dataset for the azerbaijani sign language dactyl alphabet, Speech Communication 153 (2023) 102960. doi:10.1016/j.specom.2023. 102960 on Pithdoi:10.1016/j.specom.2023
show 52 more
  1. [21] M. S. Aiswarya, R. Arockia Xavier Annie, Keyframe extraction algorithm for continuous sign-language videos using angular displacement and sequence check metrics, Interna- tionalJournalofIntelligentSystems2024(2024)4725216. doi:10.1155/2024/4725216 on Pithdoi:10.1155/2024/4725216
  2. [22] W. Hao, C. Hou, Z. Zhang, X. Zhai, L. Wang, G. Lv, A sensing data and deep learning- based sign language recognition approach, Computers and Electrical Engineering 118 (2024) 109339. doi:10.1016/j.compeleceng.2024.109339 on Pithdoi:10.1016/j.compeleceng.2024.109339
  3. [23] S. Renjith, A. Varghese, M. Rashmi, P. S.S., Transformer-based motion-visual inte- grated fusion for isolated sign language recognition, Computers and Electrical Engi- neering 130 (2026) 110902.doi:10.1016/j.compeleceng.2025.110902 on Pithdoi:10.1016/j.compeleceng.2025.110902
  4. [24] B. R, A. K. V. D, A. K. V. D, Mizo hand sign language detection using a multi- scale transformer-based hybrid feature extractor and fusion network, Computers and Electrical Engineering 131 (2026) 110916.doi:10.1016/j.compeleceng.2025.110916 on Pithdoi:10.1016/j.compeleceng.2025.110916
  5. [25] J. Wu, X. Li, T. Li, F. Meng, Y. Kong, G. Yang, L. Senhadji, H. Shu, Cslnspeech: Solving the extended speech separation problem with the help of chinese sign language, Speech Communication 165 (2024) 103131.doi:10.1016/j.specom.2024.103131. 28 on Pithdoi:10.1016/j.specom.2024.103131
  6. [26] F. M. Talaat, W. El-Shafai, N. F. Soliman, A. D. Algarni, F. E. Abd El-Samie, A. I. Siam, Real-time arabic avatar for deaf-mute communication enabled by deep learning sign language translation, Computers and Electrical Engineering 119 (2024) 109475. doi:10.1016/j.compeleceng.2... on Pithdoi:10.1016/j.compeleceng.2024.109475
  7. [27] N. C. Camgoz, O. Koller, S. Hadfield, R. Bowden, Sign language transformers: Joint end-to-end sign language recognition and translation, in: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), IEEE, Seattle, WA, USA, 2020, pp. 10020––10030. doi:10.1109/... on Pithdoi:10.1109/cvpr42600.2020.01004
  8. [28] L. Chaudhary, T. Ananthanarayana, E. Hoq, I. Nwogu, Signnet ii: A transformer-based two-way sign language translation model, IEEE Transactions on Pattern Analysis and Machine Intelligence 45 (11) (2023) 12896–12907.doi:10.1109/TPAMI.2022.3232389 on Pithdoi:10.1109/tpami.2022.3232389
  9. [29] J. Hao, H. Pan, A novel deep transformer based cvt model for sign language recog- nition in visual communication, Scientific Reports 16 (1) (2026) 1823.doi:10.1038/ s41598-025-31558-1 on Pith
  10. [30] W. F. Maia, A. M. Lopes, S. A. David, Automatic sign language to text translation using mediapipe and transformer architectures, Neurocomputing 642 (2025) 130421. doi:10.1016/j.neucom.2025.130421 on Pithdoi:10.1016/j.neucom.2025.130421
  11. [31] H. Jin, S. Zhang, Z. Yang, Q. Han, J. Cao, Exploiting locality sensitive hashing - clustering and gloss feature for sign language production, Speech Communication 171 (2025) 103227. doi:10.1016/j.specom.2025.103227 on Pithdoi:10.1016/j.specom.2025.103227
  12. [32] M. Karim, M. Ali, M. F. Hasan, M. Fatema-Tuj-Jahra, S. S. Sultana, D. M. Farid, Bangla sign language recognition using yolov5, in: 2023 IEEE 8th International Con- ference for Convergence in Technology (I2CT), IEEE, Lonavla, India, 2023, pp. 1––7. doi:10.1109/I2CT57861.2023.10126419 on Pithdoi:10.1109/i2ct57861.2023.10126419
  13. [33] M. S. U. R. Khan, P. Mehta, A. Sankar, U. Kumaravelan, S. Doddapaneni, S. B, V. G, S. Jain, A. Kunchukuttan, P. Kumar, R. Dabre, M. Khapra, Indicllmsuite: A blueprint for creating pre-training and fine-tuning datasets for indian languages, in: Proceedings of the 62nd Annual Me... on Pithdoi:10.18653/v1/2024.acl-long.843
  14. [34] J. Abadji, P. Ortiz Suarez, L. Romary, B. Sagot, Towards a cleaner document-oriented multilingual crawled corpus, in: Proceedings of the Thirteenth Language Resources and Evaluation Conference, European Language Resources Association, Marseille, France, 2022, pp. 4344–4355. do... on PithPith's reviewarXivdoi:10.48550/arxiv.2201.06642
  15. [35] J. Dodge, M. Sap, A. Marasović, W. Agnew, G. Ilharco, D. Groeneveld, M. Mitchell, M. Gardner, Documenting large webtext corpora: A case study on the colossal clean crawled corpus, in: Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, Assoc... on Pithdoi:10.18653/v1/2021
  16. [36] I. Johnson, L.-A. Kaffee, M. Redi, Wikimedia data for ai: a review of wikimedia datasets fornlptasksandai-assistedediting, in: ProceedingsoftheFirstWorkshoponAdvancing Natural Language Processing for Wikipedia, Association for Computational Linguistics, Miami, Florida, USA, 20... on Pithdoi:10.18653/v1/2024.wikinlp-1.14
  17. [38] A. Bhattacharjee, T. Hasan, W. U. Ahmad, R. Shahriyar, Banglanlg and banglat5: Benchmarks and resources for evaluating low-resource natural language generation in bangla, in: Findings of the Association for Computational Linguistics: EACL 2023, Association for Computational Li... on Pithdoi:10.18653/v1/2023.findings-eacl.54
  18. [39] A. Kilgarriff, V. Baisa, J. Bušta, M. Jakubíček, V. Kovář, J. Michelfeit, P. Rychlý, V. Suchomel, The sketch engine: ten years on, Lexicography 1 (1) (2014) 7––36.doi: 10.1007/s40607-014-0009-9 on Pithdoi:10.1007/s40607-014-0009-9
  19. [40] M. H. I. Bijoy, U. Ayman, M. M. Islam, Banglatense: A large-scale dataset of bangla sentences categorized by tense: Past, present, and future, Data in Brief 59 (2025) 111400. doi:10.1016/j.dib.2025.111400 on Pithdoi:10.1016/j.dib.2025.111400
  20. [41] T. S. Apon, R. Anan, E. A. Modhu, A. Suter, I. J. Sneha, M. G. R. Alam, Banglasarc: A dataset for sarcasm detection, in: 2022 IEEE Asia-Pacific Conference on Computer Science and Data Engineering (CSDE), IEEE, Gold Coast, Australia, 2022, pp. 1––5. doi:10.1109/CSDE56538.2022.10089322 on Pithdoi:10.1109/csde56538.2022.10089322
  21. [42] A. Moryossef, K. Yin, G. Neubig, Y. Goldberg, Data augmentation for sign lan- guage gloss translation, in: Proceedings of the 1st International Workshop on Auto- matic Translation for Signed and Spoken Languages (AT4SSL), Association for Machine Translation in the Americas, 20... on Pith
  22. [43] A. Bhattacharjee, T. Hasan, W. Ahmad, K. S. Mubasshir, M. S. Islam, A. Iqbal, M. S. Rahman, R. Shahriyar, Banglabert: Language model pretraining and benchmarks for low-resource language understanding evaluation in bangla, in: Findings of the Asso- ciation for Computational Lin... on Pithdoi:10.18653/v1/2022
  23. [44] S. Kobayashi, Contextual augmentation: Data augmentation by words with paradig- matic relations, in: Proceedings of the 2018 Conference of the North American Chap- ter of the Association for Computational Linguistics: Human Language Technolo- 30 gies, Volume 2 (Short Papers), ... on Pithdoi:10.18653/v1/n18-2072
  24. [45] J. Y. Jang, H. Park, S. Shin, S. Shin, B. Yoon, G. Gweon, Automatic gloss-level data augmentation for sign language translation, in: Proceedings of the Thirteenth Language Resources and Evaluation Conference, European Language Resources Association, Mar- seille, France, 2022, ... on Pith
  25. [46] N. Team, M. R. Costa-jussa, J. Cross, et al., No language left behind: Scaling human- centered machine translation, arXiv preprint arXiv:2207.04672 (2022).doi:10.48550/ arXiv.2207.04672 on PithPith's reviewarXivdoi:10.48550/arxiv.2207.04672
  26. [47] J. Gala, P. A. Chitale, A. K. Raghavan, V. Gumma, S. Doddapaneni, A. K. M, J. A. Nawale, A. Sujatha, R. Puduppully, V. Raghavan, P. Kumar, M. M. Khapra, R. Dabre, A. Kunchukuttan, Indictrans2: Towards high-quality and accessible machine transla- tion models for all 22 schedule... on PithPith's reviewarXivdoi:10.48550/arxiv.2305.16307
  27. [48] R. Sennrich, B. Haddow, A. Birch, Improving neural machine translation models with monolingual data, in: Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Berlin, Germany, 2016, pp. 86–96. doi:10.18653/v1/P16-1009 on Pithdoi:10.18653/v1/p16-1009
  28. [49] L. Team, The llama 3 herd of models, arXiv preprint arXiv:2407.21783 (2024).doi: 10.48550/arXiv.2407.21783 on PithPith's reviewarXivdoi:10.48550/arxiv.2407.21783
  29. [50] Y. Wang, Y. Kordi, S. Mishra, A. Liu, N. A. Smith, D. Khashabi, H. Hajishirzi, Self- instruct: Aligning language models with self-generated instructions, in: Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Toront... on Pithdoi:10.18653/v1/2023
  30. [51] C. Whitehouse, M. Choudhury, A. F. Aji, LLM-powered data augmentation for en- hanced cross-lingual performance, in: Proceedings of the 2023 Conference on Empir- ical Methods in Natural Language Processing, Singapore, 2023, pp. 671–686. doi: 10.18653/v1/2023.emnlp-main.44 on Pithdoi:10.18653/v1/2023.emnlp-main.44
  31. [52] S. Węglarczyk, Kernel density estimation and its application, ITM Web of Conferences 23 (2018) 00037. doi:10.1051/itmconf/20182300037 on Pithdoi:10.1051/itmconf/20182300037
  32. [53] L. McInnes, J. Healy, N. Saul, L. Großberger, Umap: Uniform manifold approximation and projection, Journal of Open Source Software 3 (29) (2018) 861.doi:10.21105/ joss.00861 on Pith
  33. [54] D. C. van Leijenhorst, T. P. van der Weide, A formal derivation of heaps’ law, Infor- mation Sciences 170 (2–4) (2005) 263––272.doi:10.1016/j.ins.2004.03.006. 31 on Pithdoi:10.1016/j.ins.2004.03.006
  34. [55] M. Lewis, Y. Liu, N. Goyal, M. Ghazvininejad, A. Mohamed, O. Levy, V. Stoyanov, L. Zettlemoyer, Bart: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension, in: Proceedings of the 58th Annual Meet- ing of the Association fo... on Pithdoi:10.18653/v1/2020.acl-main.703
  35. [56] A. CONNEAU, G. Lample, Cross-lingual language model pretraining, in: H. Wallach, H. Larochelle, A. Beygelzimer, F. d'Alché-Buc, E. Fox, R. Garnett (Eds.), Advances in Neural Information Processing Systems, Vol. 32, Curran Associates, Inc., 2019.doi: 10.48550/arXiv.1901.07291 on PithPith's reviewarXivdoi:10.48550/arxiv.1901.07291
  36. [57] E. J. Hu, Y. Shen, P. Wallis, Z. Allen-Zhu, Y. Li, S. Wang, L. Wang, W. Chen, LoRA: Low-rank adaptation of large language models, in: The Tenth International Conference on Learning Representations, Virtual, 2022.doi:10.48550/arXiv.2106.09685 on PithPith's reviewarXivdoi:10.48550/arxiv.2106.09685
  37. [58] Y. Liu, J. Gu, N. Goyal, X. Li, S. Edunov, M. Ghazvininejad, M. Lewis, L. Zettlemoyer, Multilingual denoising pre-training for neural machine translation, Transactions of the Association for Computational Linguistics 8 (2020) 726––742.doi:10.1162/tacl_a_ 00343 on Pithdoi:10.1162/tacl_a_
  38. [59] L. Xue, N. Constant, A. Roberts, M. Kale, R. Al-Rfou, A. Siddhant, A. Barua, C. Raf- fel, mt5: A massively multilingual pre-trained text-to-text transformer, in: Proceedings of the 2021 Conference of the North American Chapter of the Association for Com- putational Linguistics... on Pithdoi:10.18653/v1/2021.naacl-main.41
  39. [60] M. S. Salim, H. Murad, D. Das, F. Ahmed, Banglagpt: A generative pretrained transformer-based model for bangla language, in: 2023 International Conference on In- formation and Communication Technology for Sustainable Development (ICICT4SD), IEEE, Dhaka, Bangladesh, 2023, pp. 5... on Pithdoi:10.1109/icict4sd59951.2023
  40. [61] C.-Y. Lin, ROUGE: A package for automatic evaluation of summaries, in: Text Sum- marization Branches Out, Association for Computational Linguistics, Barcelona, Spain, 2004, pp. 74–81 on Pith
  41. [62] K. Papineni, S. Roukos, T. Ward, W.-J. Zhu, Bleu: a method for automatic evaluation of machine translation, in: Proceedings of the 40th Annual Meeting on Association for Computational Linguistics - ACL ’02, Association for Computational Linguistics, Philadelphia, Pennsylvania,... on PithPith's reviewdoi:10.3115/1073083.1073135
  42. [63] R. Rei, C. Stewart, A. C. Farinha, A. Lavie, Comet: A neural framework for mt eval- uation, in: Proceedings of the 2020 Conference on Empirical Methods in Natural Lan- guage Processing (EMNLP), Association for Computational Linguistics, Online, 2020, pp. 2685––2702. doi:10.186... on Pithdoi:10.18653/v1/2020.emnlp-main.213
  43. [64] D. Biderman, J. Portes, J. J. G. Ortiz, M. Paul, P. Greengard, C. Jennings, D. King, S. Havens, V. Chiley, J. Frankle, C. Blakeney, J. P. Cunningham, LoRA learns less and 32 forgets less, Transactions on Machine Learning Research (2024).doi:10.48550/arXiv. 2405.09673 on PithPith's reviewarXivdoi:10.48550/arxiv.2405.09673
  44. [65] R. Shuttleworth, J. Andreas, A. Torralba, P. Sharma, Lora vs full fine-tuning: An illusion of equivalence, in: Advances in Neural Information Processing Systems 38, Neural Information Processing Systems Foundation, Inc. (NeurIPS), San Diego, Cal- ifornia, USA and Mexico City, ... on Pith
  45. [66] T. Su, X. Peng, S. Thillainathan, D. Guzmán, S. Ranathunga, E.-S. Lee, Unlock- ing parameter-efficient fine-tuning for low-resource language translation, in: Find- ings of the Association for Computational Linguistics: NAACL 2024, Association for Computational Linguistics, Mexi... on Pithdoi:10.18653/v1/2024.findings-naacl.263
  46. [67] Y. Zeng, K. Lee, The expressive power of low-rank adaptation, in: The Twelfth Inter- national Conference on Learning Representations, OpenReview.net, Vienna, Austria, on Pith
  47. [68] S. M. Abdullah, A. Paul, S. Rayana, A. Kabir, Z. Masud, State-of-the-art translation of text-to-gloss using mbart : A case study of bangla (2025). arXiv:2504.02293, doi:10.48550/arXiv.2504.02293 on PithPith's reviewarXivdoi:10.48550/arxiv.2504.02293
  48. [69] A. Duarte, S. Palaskar, L. Ventura, D. Ghadiyaram, K. DeHaan, F. Metze, J. Tor- res, X. Giro-i Nieto, How2sign: A large-scale multimodal dataset for continuous amer- ican sign language, in: 2021 IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR), IEEE, Nas... on Pithdoi:10.1109/cvpr46437.2021.00276
  49. [70] A. Joshi, R. Mohanty, M. Kanakanti, A. Mangla, S. Choudhary, M. Barbate, A. Modi, isign: A benchmark for indian sign language processing, in: Findings of the Association for Computational Linguistics ACL 2024, Association for Computational Linguistics, Bangkok, Thailand and vi... on Pithdoi:10.18653/v1/
  50. [71] H. A. Rubaiyeat, H. Mahmud, M. K. Hasan, Bangla sign language translation: Dataset creation challenges, benchmarking and prospects (2025).arXiv:2511.21533 on PitharXiv
  51. [72] N. Saha, R. Shahriyar, N. A. Roudra, S. Sakib, A. A. Rasel, Introducing a Bangla sentence – gloss pair dataset for Bangla sign language translation and research, in: Proceedings of the Fifteenth Language Resources and Evaluation Conference, ELRA Language Resource Association, ... on Pithdoi:10.63317/38qenrwzegr9
  52. [2024] doi:10.48550/arXiv.2310.17513 on PithPith's reviewarXivdoi:10.48550/arxiv.2310.17513

Challenges and discussion

Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.

No comments yet.

Sign in to comment