NushuRescue: Revitalization of the Endangered Nushu Language with AI

Ivory Yang; Soroush Vosoughi; Weicheng Ma

REVIEW 5 major objections 6 minor 12 references

NushuRescue: Revitalization of the Endangered Nushu Language with AI

T0 review · 5 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read An LLM with no prior exposure to Nüshu learned it from 35 examples and translated 48.69% of held-out sentences exactly.

desk verdict The dataset is a real contribution, but the headline 48.69% accuracy cannot be a perfect-match count on 50 sentences, so the central number needs checking before this is accepted. read the letter →

arxiv 2412.00218 v4 pith:Y4NXGNR2 submitted 2024-11-29 cs.CL cs.LG

Ivory Yang , Weicheng Ma , Soroush Vosoughi This is my paper · ORCID

classification cs.CLcs.LG

keywords Nüshuendangeredlanguagepreservationlow-resourceNLPfew-shotmachinetranslationlargemodelsparallelcorpusconstructiondataaugmentationwomen'sscript

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 claims that a general-purpose large language model with no prior knowledge of Nüshu can learn to translate into this endangered syllabic script from just 35 example sentences, reaching 48.69% character-exact accuracy on 50 held-out sentences. That result matters because Nüshu has almost no digital resources: the only expert-validated source is a print compendium, and manual transcription is slow and costly. The paper's NüshuRescue framework pairs few-shot prompting with an automated length validator and iterative corpus expansion, producing a 500-sentence gold corpus (NCGold) and a 98-sentence machine-generated extension (NCSilver). If the claim holds, AI-assisted data generation can lower the human cost of revitalizing other low-resource languages.

What carries the argument

The load-bearing mechanism is the NüshuRescue few-shot generation loop: the LLM receives uploaded dictionary data and 35 seed Nüshu–Chinese pairs, then translates new Chinese sentences, while a rule-based validator rejects any output whose Nüshu character count does not match the Chinese source and prompts up to seven retries. Across six rounds, the top five newly generated translations replace the bottom five seed examples, gradually extending to longer sentences. The NCGold corpus—500 manually transcribed pairs from a printed Nüshu compendium, including 373 sentences that required manual visual mapping beyond the official dictionary—supplies the seed data and the evaluation set that anchors the claimed accuracy.

What would settle it

Re-run the 35-example few-shot prompt with the Nüshu symbols in the seed pairs randomly permuted, keeping the Chinese sides fixed, then score the 50 withheld sentences on exact character match; if accuracy stays near 48.69%, the result is carried by the length validator and restricted vocabulary rather than learned Nüshu-to-Chinese mappings, while a collapse toward chance would support genuine script acquisition.

Watch

Extended reading notes

Core claim

The central discovery is that the knowledge needed to translate into a truly low-resource script can be supplied almost entirely at inference time: GPT-4-Turbo, shown to be unable to recognize Nüshu before prompting, achieved 48.69% exact character-match accuracy on 50 withheld NCGold sentences after seeing only 35 seed pairs plus dictionary files. The paper attributes most errors to Nüshu's one-to-many and many-to-one character mappings, where multiple Chinese homophones share a syllable. Enforcing a one-to-one character-length rule raised accuracy from 31.37% to 48.69%, showing that language-specific structural constraints can partly compensate for missing data. The same pipeline generated 98 new Chinese-to-Nüshu translations (NCSilver) and, applied to Cherokee with larger seed sets, produced 28%, 40%, and 42% exact accuracy across seed sizes of 35, 150, and 300, suggesting the method transfers beyond Nüshu.

Load-bearing premise

Everything rests on NCGold being accurate: its annotators were bilingual and computationally trained but not Nüshu experts, and for 373 of 500 sentences they had to visually match characters outside the official dictionary, so a systematic mapping error would make the claimed accuracy a comparison against a flawed benchmark.

Editorial extensions

If this is right

Human effort for endangered-language corpus building drops from full manual transcription to producing a few dozen validated seed pairs.
The resulting NCGold and NCSilver datasets give the research community the first public Nüshu parallel corpus, covering 80% of the only expert-validated Nüshu source.
Because the LLM needs no prior exposure to the target script, the framework can be pointed at other undocumented or underdocumented languages immediately, limited mainly by seed data.
Enforcing simple script-specific constraints, such as the one-to-one character-length rule, can substitute for large training sets in low-resource translation.
The Seq2Seq experiments imply that LLM-generated data continues to improve downstream translation models, up to a length and complexity threshold beyond which performance declines.

Reading between the lines

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

The paper leaves an expert validation pass as future work; if such a pass corrected even a modest share of the 50 withheld NCGold sentences, the 48.69% figure would need recomputation, so the headline number is only as solid as the annotators' manual visual matching.
Because the last native Nüshu speaker died in 2004, any machine-generated revival text enters circulation without native-speaker checks; a deployment pipeline would need community or expert sign-off before out-of-dictionary generated forms are used pedagogically.
The no-prior-exposure property suggests the pipeline could be pointed at orally attested but unwritten languages, shifting the bottleneck from data volume to trusted seed annotation by community partners.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

First, the good news: NCGold and NCSilver are genuinely new and useful artifacts. There was no public Nushu-Chinese parallel corpus before, and this one comes with code and a documented pipeline. The few-shot prompting setup with the length validator is a sensible adaptation of in-context learning to a nearly-zero-resource script, and the Cherokee experiments are a nice sanity check that the approach transfers. The paper also deserves credit for being upfront about the biggest limitation: without a native-expert check, they won't release out-of-dictionary generations.

But the headline number does not survive contact with arithmetic. 48.69% accuracy on exactly 50 withheld sentences means 24.345 perfect matches. That is impossible if the metric is a sentence-level perfect-match count, which is exactly how Section 4.2.3 describes it. Either the metric is actually a character-level average, the test set is not 50, or the number is misreported. On top of that, the paper says the authors manually reviewed and 'corrected minor character choice errors' before reporting that figure, so it is unclear whether 48.69% is the raw model output or output after human correction. As stated, the central quantitative claim is not reproducible.

The second soft spot is the gold standard itself. The annotators are bilingual and computationally literate, but they are not Nushu experts, and 373 of 500 sentences required visual matching outside the official dictionary. The introduction and Section 4.1 call NCGold 'expert-validated,' but the Limitations section says the lack of expert validation is exactly why generated data could not be released. That is an internal contradiction, and it matters because every downstream number—including the 48.69%—is measured against this reference. If those 373 manual mappings contain systematic errors, the headline is measured against a corrupted standard.

To be clear, I am not saying the framework is a failure. The qualitative result—that an off-the-shelf LLM can learn a syllabic script from 35 examples—is plausible and worth testing properly. The dataset alone is a contribution. But the paper as written overstates certainty: the abstract and conclusion call the translations 'high-quality' and 'expert-validated' without support. The fix is straightforward: report the exact evaluation protocol (sentence-level vs. character-level), give per-round counts and confidence intervals, clarify what the human correction step did, and either get partial expert validation for NCGold or soften the 'expert-validated' language.

This is a legitimate paper for a serious venue, but it needs a major revision focused on the evaluation. I would send it to peer review with a clear request to resolve the arithmetic and the gold-standard validation. The stress-test note is right on both counts.

Referee Report

5 major / 6 minor

Summary. The paper introduces NüshuRescue, an LLM-in-the-loop framework for scaling up corpora of endangered languages with minimal human effort, and contributes two new resources: NCGold, a 500-sentence Nüshu–Chinese parallel corpus, and NCSilver, 98 LLM-generated Nüshu–Chinese sentence pairs. The central empirical claim is that GPT-4-Turbo, given only 35 seed pairs from NCGold and a character-count length validator, achieves 48.69% strict character-level translation accuracy on 50 withheld NCGold sentences. The paper also trains a FastText language model and a Seq2Seq Nüshu-to-Chinese translation model on the resulting corpora, reporting that translation metrics improve as the training set grows. The authors include a Limitations section that acknowledges the absence of expert validation and explains why out-of-dictionary generated data was not released.

Significance. If the reported numbers hold, the paper provides a valuable public dataset for a truly endangered language, a simple and reproducible data-augmentation pipeline, and transfer evidence on Cherokee, all of which would be useful for the low-resource NLP community. The release of code and data, the documentation of prompts, and the unusually candid Limitations section are strengths. However, the headline accuracy is measured against a gold standard whose expert validation is asserted in some places and denied in others, and the reported percentage is inconsistent with the stated evaluation size. The significance is therefore conditional: the resource and framework are promising, but the central quantitative claim needs to be either corrected or substantially re-framed before the paper can be accepted.

major comments (5)

[Abstract; §4.2.3] The abstract and §4.2.3 report 48.69% accuracy on 50 withheld sentences under a strict sentence-level character-by-character match. Since 48.69% of 50 is 24.345, this number cannot be a count of perfectly translated sentences. Either the metric is an average over a different unit (for example, per-character accuracy), the evaluation set does not actually contain 50 sentences, or the number is misreported. Please give the raw counts and define the metric unambiguously; the headline result must be reproducible from the reported numbers.
[§4.1; Limitations] The manuscript repeatedly calls NCGold 'expert-validated' (Abstract, §1, §4.1), but §4.1 states that the annotators are bilingual with computational linguistics experience, not Nüshu experts, and that 373 of the 500 sentences required manual visual matching outside the official dictionary. The Limitations section then states that 'the absence of expert validation prevents us from safely releasing the generated data.' These statements are in direct tension. If the 373 manually mapped sentences contain systematic character-selection errors, the 48.69% accuracy, the NCSilver examples, and all Seq2Seq evaluations are measured against a flawed reference. Please state clearly which portion of NCGold is expert-validated, describe the validation procedure actually used, and report agreement or an error audit.
[§4.2.1; §4.2.3] The evaluation protocol is ambiguous about whether the reported accuracy is computed on raw GPT-4-Turbo outputs or on outputs after the authors' manual review that 'corrected minor character choice errors.' This distinction is load-bearing because the central claim is about the model's capability, not about human post-editing. Please report raw and post-correction accuracy separately, and specify how disagreements were adjudicated.
[§3; §4.2.2] The length validator is described as ensuring that 'each Nüshu character maps to a single Chinese character,' yet §3 explicitly describes one-to-many and many-to-one mappings between Nüshu and Chinese, with one Nüshu word mapping to up to 35 Chinese words. The equal-length constraint is therefore not generally guaranteed for correct translations. Using this validator to filter generations could bias the evaluation by rejecting valid translations or accepting incorrect ones solely on length. Please justify the one-to-one length assumption on the actual NCGold data and quantify how many seed and test sentences satisfy it.
[§4.2.1; §4.2.2] The number of seed examples (35) and the length-validation rule are said to be informed by 'preliminary tests,' but the paper does not state whether those tests involved the 50 withheld sentences or otherwise influenced design choices on the evaluation set. If any tuning used the evaluation set, the 48.69% figure is not a clean held-out estimate. In addition, a 50-sentence test set at roughly 49% accuracy has a binomial standard error of about 7 percentage points; please report bootstrap or exact binomial confidence intervals and state the precise train/test split.

minor comments (6)

[Throughout] The spelling of the language name is inconsistent: the title and most sections use 'Nüshu,' but the abstract and several places use 'Nushu.' Please standardize.
[§4.2.2; Figure 10] The stacked bar chart in Figure 10 is difficult to read because the count labels are small and partially overlapping; please add numeric labels in a table or use a clearer chart format so that the per-round counts and control-group counts can be verified.
[§5.2; Table 2] Table 2 does not state whether the fixed 100-sentence test set is drawn from NCGold only or from the combined NCGold+NCSilver corpus, nor how the test set was selected after shuffling. Please clarify this in the text and table caption.
[§6] The claim that 'the maximum sentence length in the original 500-sentence dataset is only 18' needs verification and a supporting length distribution, especially since Table 1 reports average lengths only for the generated rounds and the original NCGold corpus is described as containing poetry and prose of varying lengths.
[§4.2.3] The statement that GPT-4-Turbo successfully translated '70.00% of sentences with lengths below 16 characters' should include the raw numerator and denominator, as well as the exact length bin definition, so that the reader can assess the reliability of this secondary result.
[Limitations; footnote 5] The factual claim that the last known native speaker of Nüshu passed in 2004 is given without a citation; please provide a reference for this important contextual statement.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the headline 48.69% is a measured held-out accuracy, not a quantity forced by construction or by self-citation.

full rationale

The central quantitative claim—48.69% strict character-level accuracy on 50 withheld NCGold sentences after few-shot prompting with 35 NCGold seeds—is an empirical held-out measurement, not a quantity derived from its own inputs by construction. The 35 seed sentences and the supplied Nüshu–Chinese dictionary are inputs; the 50 withheld gold translations are not included in the prompt, and the length validator constrains only output length, not character identity, so matching the gold reference remains a genuinely measured event. The Seq2Seq experiments likewise train on NCGold/NCSilver and evaluate on a fixed 100-sentence NCGold test set; no test-set values are fitted or renamed as predictions. No load-bearing self-citation chain appears: the paper's authoritative source (Zhao 1992's Compendium) and the official Nüshu Unicode dictionary are external to the authors. Concerns about nonexpert annotation, unvalidated NCSilver, or the fractional 48.69% count bear on data quality and reproducibility, not on circular derivation.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central result rests on the quality of NCGold and on the one-to-one length assumption. Both are unvalidated by native Nushu experts. The headline accuracy is a measurement, not a derivation.

free parameters (5)

Number of seed examples = 35
Chosen after preliminary tests where fewer examples made the model refuse; this hand-tuned number directly sets the few-shot condition for the headline accuracy.
Length-validated one-to-one character mapping = 1 Nushu character = 1 Chinese character
Enforced by the rule-based validator; the paper reports this raises accuracy from 31.37% to 48.69%, so the headline result depends on this chosen constraint.
Per-round replacement count = 5
In rounds 2+, five newly generated samples replace five of the 35 seed examples; chosen without reported ablations.
Maximum retries for failed translations = 7
Pipeline stops after 7 retries and marks failure; this threshold affects NCSilver size and success rates.
Seq2Seq architecture hyperparameters = batch=3, epochs=15, embedding=300, window=10, min_count=5, ns=10, lr=0.05
Hand-set for the FastText and Seq2Seq models; not central to the LLM claim but determine the reported BLEU and ROUGE numbers in Table 2.

assumptions (5)

ad hoc to paper Nushu-Chinese mapping is one-to-one in character count for in-dictionary sentences.
The length validator rejects any translation whose Nushu length differs from the Chinese input, which is restrictive given the paper's own description of one-to-many and many-to-one mappings in Section 3.
domain assumption The source book 'A Compendium of Chinese Nushu' contains accurate expert-validated translations.
The entire NCGold corpus is transcribed from this book; no independent expert verification is performed on the digitized result (Section 4.1).
domain assumption The official Nushu-Chinese Unicode dictionary and online converter provide correct mappings.
Used as the primary mapping tool; 373 of 500 sentences required manual visual matching beyond dictionary coverage, so correctness of this tool is load-bearing.
ad hoc to paper GPT-4-Turbo had no prior exposure to Nushu.
Claimed in the abstract and Section 4.2, but the model's training data are undisclosed; an initial failure example is used as evidence rather than a verifiable property.
ad hoc to paper Annotators with computational linguistics experience and Chinese-English bilingualism can reliably transcribe Nushu.
No native or expert Nushu speaker is involved; the limitation section concedes that absence of expert validation prevents releasing out-of-dictionary generated data.

how reviews work

0 comments

Cite this review

Pith. "Pith review of NushuRescue: Revitalization of the Endangered Nushu Language with AI." pith.science (2026). https://pith.science/paper/Y4NXGNR2

@misc{pith2026241200218,
  author       = {Pith},
  title        = {Pith review of: NushuRescue: Revitalization of the Endangered Nushu Language with AI},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y4NXGNR2}},
  note         = {Machine review of arXiv:2412.00218}
}

read the original abstract

The preservation and revitalization of endangered and extinct languages is a meaningful endeavor, conserving cultural heritage while enriching fields like linguistics and anthropology. However, these languages are typically low-resource, making their reconstruction labor-intensive and costly. This challenge is exemplified by Nushu, a rare script historically used by Yao women in China for self-expression within a patriarchal society. To address this challenge, we introduce NushuRescue, an AI-driven framework designed to train large language models (LLMs) on endangered languages with minimal data. NushuRescue automates evaluation and expands target corpora to accelerate linguistic revitalization. As a foundational component, we developed NCGold, a 500-sentence Nushu-Chinese parallel corpus, the first publicly available dataset of its kind. Leveraging GPT-4-Turbo, with no prior exposure to Nushu and only 35 short examples from NCGold, NushuRescue achieved 48.69% translation accuracy on 50 withheld sentences and generated NCSilver, a set of 98 newly translated modern Chinese sentences of varying lengths. A sample of both NCGold and NCSilver is included in the Supplementary Materials. Additionally, we developed FastText-based and Seq2Seq models to further support research on Nushu. NushuRescue provides a versatile and scalable tool for the revitalization of endangered languages, minimizing the need for extensive human input.

Figures

Figures reproduced from arXiv: 2412.00218 by the authors.

**Figure 1.** A simplified, stylized rendition of the NüshuRescue generation framework, using samples from NCGold. 2013). In a past era of high female illiteracy, where women’s voices were largely suppressed (Leung, 2003), Nüshu served as a private means of communication and form of female empowerment in a male-dominated society (Liu, 1997). Today, Nüshu is an endangered language with very limited surviving documentation (Luo e… view at source ↗

**Figure 2.** Comparison of mapped Chinese (logographic) [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** Nüshu-Chinese one-to-many and many-to-one [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

Figures from the paper (6 more)

**Figure 4.** Figure 4: Manual Data Collection: Step-by-step tran [PITH_FULL_IMAGE:figures/full_fig_p003_4.png]

**Figure 6.** Figure 6: GPT-4-Turbo does not understand Nüshu at [PITH_FULL_IMAGE:figures/full_fig_p004_6.png]

**Figure 8.** Figure 8: Pipeline for LLM-in-the-loop data annotation [PITH_FULL_IMAGE:figures/full_fig_p005_8.png]

**Figure 9.** Figure 9: Translation of unseen characters (boxed in [PITH_FULL_IMAGE:figures/full_fig_p005_9.png]

**Figure 10.** Figure 10: Translation success and failure across rounds. [PITH_FULL_IMAGE:figures/full_fig_p006_10.png]

**Figure 11.** Figure 11: Out-of-domain examples; for the incorrect [PITH_FULL_IMAGE:figures/full_fig_p006_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

12 extracted references · 8 canonical work pages

[7]

arXiv preprint arXiv:2411.11295

Tran- scending language boundaries: Harnessing llms for low-resource language translation. arXiv preprint arXiv:2411.11295. Cathy Lyn Silber

arXiv
[9]

In ACM SIGGRAPH Asia 2023 Art Gallery, pages 1–2

Ai nüshu (women’s scripts)-an ex- ploration of language emergence in sisterhood. In ACM SIGGRAPH Asia 2023 Art Gallery, pages 1–2. Thomas Thurnell-Read, Michael Skey, and Marie Heˇr- manová

work page 2023
[11]

In Proceedings of the 2020 Conference on Empirical Methods in Natu- ral Language Processing (EMNLP), pages 577–595

Chren: Cherokee-english machine translation for en- dangered language revitalization. In Proceedings of the 2020 Conference on Empirical Methods in Natu- ral Language Processing (EMNLP), pages 577–595. Liming Zhao

work page 2020
[1992]

Tsinghua University Press, Beijing

A Compendium of Chinese Nüshu (Zhongguo Nüshu Jicheng) , original print edition, re-released as an ebook in 2019 edition. Tsinghua University Press, Beijing. Liming Zhao

work page 2019
[2012]

Nüshu Script Contributors

Heroines of jiangyong: Chinese narrative ballads in women’s script. Nüshu Script Contributors. Nüshu-chinese unicode dictionary. https://github.com/nushu-script/ unicode_nushu. Accessed: 2024-09-15. Nüshu Script Online Converter. Converter of calligra- phy copybook of nüshu standard characters. https: //nushuscript.org/en-US/unicode/. Accessed: 2024-09-15...

work page 2024
[2013]

arXiv preprint arXiv:1301.3781

Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781. Riccardo Moratto

arXiv
[2018]

arXiv preprint arXiv:1806.02901

Probabilistic fasttext for multi-sense word embeddings. arXiv preprint arXiv:1806.02901. Rupjyoti Baruah, Rajesh Kumar Mundotiya, and Anil Kumar Singh

arXiv
[2019]

Unsung Challenges of Building and Deploying Language Technologies for Low Resource Language Communities

Unsung challenges of building and deploying language technologies for low resource language communities. arXiv preprint arXiv:1912.03457. Panteleimon Krasadakis, Evangelos Sakkopoulos, and Vassilios S Verykios

work page Pith review arXiv 1912

Show all 12 references

[2020]

arXiv preprint arXiv:2006.07264

Low-resource languages: A review of past work and future challenges. arXiv preprint arXiv:2006.07264. Tomas Mikolov, Kai Chen, Greg Corrado, and Jef- frey Dean

2006 arXiv
[2021]

Accessed: 2024-09-15

Universal dependencies: Chinese gsd (simplified, ud version 2.9). Accessed: 2024-09-15. Xiaobo Wang

2024
[2023]

In SIG- GRAPH Asia 2023 Art Papers, pages 1–7

Ai nüshu: An exploration of language emergence in sisterhood through the lens of computational linguistics. In SIG- GRAPH Asia 2023 Art Papers, pages 1–7. Li Hai Tan, Ho-Ling Liu, Charles A Perfetti, John A Spinks, Peter T Fox, and Jia-Hong Gao

2023
[2024]

InPro- ceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 4368–

Data collection pipeline for low-resource languages: A case study on constructing a tetun text corpus. InPro- ceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 4368–

2024

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [7]

arXiv preprint arXiv:2411.11295

Tran- scending language boundaries: Harnessing llms for low-resource language translation. arXiv preprint arXiv:2411.11295. Cathy Lyn Silber

arXiv

[2] [9]

In ACM SIGGRAPH Asia 2023 Art Gallery, pages 1–2

Ai nüshu (women’s scripts)-an ex- ploration of language emergence in sisterhood. In ACM SIGGRAPH Asia 2023 Art Gallery, pages 1–2. Thomas Thurnell-Read, Michael Skey, and Marie Heˇr- manová

work page 2023

[3] [11]

In Proceedings of the 2020 Conference on Empirical Methods in Natu- ral Language Processing (EMNLP), pages 577–595

Chren: Cherokee-english machine translation for en- dangered language revitalization. In Proceedings of the 2020 Conference on Empirical Methods in Natu- ral Language Processing (EMNLP), pages 577–595. Liming Zhao

work page 2020

[4] [1992]

Tsinghua University Press, Beijing

A Compendium of Chinese Nüshu (Zhongguo Nüshu Jicheng) , original print edition, re-released as an ebook in 2019 edition. Tsinghua University Press, Beijing. Liming Zhao

work page 2019

[5] [2012]

Nüshu Script Contributors

Heroines of jiangyong: Chinese narrative ballads in women’s script. Nüshu Script Contributors. Nüshu-chinese unicode dictionary. https://github.com/nushu-script/ unicode_nushu. Accessed: 2024-09-15. Nüshu Script Online Converter. Converter of calligra- phy copybook of nüshu standard characters. https: //nushuscript.org/en-US/unicode/. Accessed: 2024-09-15...

work page 2024

[6] [2013]

arXiv preprint arXiv:1301.3781

Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781. Riccardo Moratto

arXiv

[7] [2018]

arXiv preprint arXiv:1806.02901

Probabilistic fasttext for multi-sense word embeddings. arXiv preprint arXiv:1806.02901. Rupjyoti Baruah, Rajesh Kumar Mundotiya, and Anil Kumar Singh

arXiv

[8] [2019]

Unsung Challenges of Building and Deploying Language Technologies for Low Resource Language Communities

Unsung challenges of building and deploying language technologies for low resource language communities. arXiv preprint arXiv:1912.03457. Panteleimon Krasadakis, Evangelos Sakkopoulos, and Vassilios S Verykios

work page Pith review arXiv 1912