Pith. sign in

REVIEW 4 major objections 4 minor 1 cited by

RAIR: Retrieval-Augmented Iterative Refinement for Chinese Spelling Correction

T0 review · 4 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read RAIR is a plug-and-play retrieval-augmented framework that improves LLM Chinese spelling correction in specialized domains and variable-length settings without fine-tuning the LLM.

desk verdict RAIR is a plausible plug-and-play RAG framework for Chinese spelling correction, but the LEMON zero-shot protocol leaks test-side information through self-generated retrievals, undermining the headline domain-adaptation claim. read the letter →

arxiv 2504.18938 v2 pith:CYS5HV4Y submitted 2025-04-26 cs.CL

Junhong Liang , Yu Zhou This is my paper · ORCID
classification cs.CL
keywords ChineseSpellingCorrectionRetrieval-AugmentedGenerationDomainAdaptationLargeLanguageModelsMulti-turnLengthReflectionASRN-bestCharacterSplittingVariable-length
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

Chinese spelling correction usually assumes the corrected sentence has the same length as the input, and it struggles when specialized terminology appears in legal, medical, or other domain texts. This paper proposes RAIR, a retrieval-augmented framework that supplies an LLM with retrieved correct sentences and domain knowledge, then iteratively checks and repairs the output length. The authors claim RAIR beats current approaches on domain spelling benchmarks and substantially improves LLMs on variable-length tasks such as ASR N-best correction and character-splitting restoration. The payoff is a model-agnostic recipe: the same prompt-and-retrieve loop can be attached to any instruction-following LLM without retraining it.

What carries the argument

The mechanism is the retrieval-and-refine loop. A retrieval corpus is built by combining domain terms with machine-generated explanations, correct sentences from training data, and expanded paragraphs generated from those sentences; a dense retriever is fine-tuned with contrastive pairs so an erroneous query still finds its clean counterpart. Multi-turn Length Reflection is the named loop that carries the length guarantee: after each generation it compares source and output lengths, formats a length report into a prompt, and asks the LLM to regenerate, repeating until the constraint is met or a round limit is reached. Adaptive Selection routes between retrieval-augmented and direct generation, switching methods when the length check fails. Together these modules let the framework enforce equal-length correction for spelling errors and bounded-length correction for splitting and N-best errors.

What would settle it

Rebuild the LEMON retrieval corpus from domain documents and dictionaries that contain none of the test sentences, rerun DeepSeek-V3+RAIR, and compare F1; if the gap over the non-retrieval baseline mostly disappears, the reported zero-shot advantage rests on test-sentence-derived contexts.

Watch

Extended reading notes

Core claim

RAIR is a plug-and-play retrieval-augmented framework for Chinese spelling correction. It builds a retrieval corpus from three sources—domain terms with LLM-written explanations, the correct sentences in the training set, and LLM-expanded paragraphs derived from those correct sentences—then fine-tunes a dense retriever on positive and negative pairs so that it can match an error-containing query to its clean target. At inference time the framework feeds the LLM the retrieved sentences along with the source text, applies a Multi-turn Length Reflection loop that checks output length against the task's constraint and asks the model to regenerate when the check fails, and uses Adaptive Selection to choose between retrieval-based and direct correction. The paper reports that this recipe, without any LLM fine-tuning, improves F1 over the base LLM across ECSpell and LEMON domains, lowers CER on the Aishell-1 ASR N-best benchmark, and improves splitting correction on CSEC, with the largest relative gains on weaker instruction followers.

Load-bearing premise

The LEMON results assume that an LLM-generated paragraph derived from each test sentence and later retrieved back does not leak the correct answer; if that assumption fails, the zero-shot domain-spelling gains are inflated.

Editorial extensions

If this is right

  • On the ECSpell and LEMON domain benchmarks, RAIR improves F1 over every base LLM in every domain tested; DeepSeek-V3+RAIR reports the best F1 in official-document writing and all seven LEMON domains.
  • On Aishell-1 ASR N-best correction, RAIR lowers CER for all three LLMs, cutting GPT-3.5's CER by 35.9% (from 9.84% to 6.31%) and reaching 4.15% CER with DeepSeek-V3.
  • On CSEC splitting correction, RAIR raises DeepSeek-V3's F1 from 54.2 to 66.1 on news and from 47.0 to 52.0 on social text, and it is the best method on social text.
  • Ablations indicate that Multi-turn Length Reflection is the single largest contributor: removing it drops DeepSeek-V3's F1 by more than ten points on most LEMON domains, which fits the exact-match evaluation criterion.
  • Because RAIR never changes the generator's weights, the same framework ports to new backbones simply by swapping the LLM and rebuilding the retrieval corpus.

Reading between the lines

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

  • The same retrieve-and-reflect loop should transfer to other noisy-text generation tasks with hard length constraints, such as translating with character limits or extracting fixed-width records, since the length report is independent of Chinese.
  • The retriever fine-tuning recipe—matching error-containing queries to clean targets while using semantic distractors as negatives—is a general pattern for retrieval over noisy user-generated text, not just spelling.
  • The framework's adaptive selection suggests that retrieval adds the most value for weaker instruction followers; a deployment could route only low-confidence or low-capability models through the full RAG pipeline to save API cost.
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.

Referee Report

4 major / 4 minor

Summary. The paper introduces RAIR, a plug-and-play retrieval-augmented generation framework for Chinese spelling correction. RAIR constructs a retrieval corpus from domain lexicons, training sentences, and LLM-generated expansions; fine-tunes a bge-m3 retriever with contrastive samples; runs a multi-turn length-reflection (MLR) loop to enforce output-length constraints; and uses an adaptive-selection module to choose between retrieval-based and non-retrieval correction. The authors evaluate on three tasks: equal-length domain spelling correction (ECSpell and LEMON), ASR N-best correction (ChineseHP/Aishell-1), and character-splitting correction (CSEC), and report gains over several LLM backbones.

Significance. RAIR addresses a real gap: LLM-based CSC often fails on domain terms and variable-length outputs, and a model-agnostic wrapper that enforces length via reflection is a practical contribution. The paper's strengths include clear framing of equal- and variable-length error types, a sensible use of external lexicons for ECSpell, and a useful ablation isolating MLR and AdSe. The MLR results are internally consistent and show large gains in length-constrained settings. However, the central claim that RAIR 'outperforms current approaches in domain spelling correction' rests heavily on LEMON, whose zero-shot retrieval corpus is built from the test sentences themselves; that evaluation is not leakage-free. The comparison set also omits the most relevant recent RAG-based and LLM-based CSC systems, so the SOTA claim is not established.

major comments (4)
  1. [Section 3.2, LEMON protocol] The LEMON retrieval corpus is constructed by generating a background description with GPT-4o for each test source sentence and then using that description as the retrieval result for the same sentence. The claim that this 'ensures no data leakage occurs' is incorrect: the retrieved context is derived from the very input being corrected, and GPT-4o frequently repairs or paraphrases the error in the description, so the retrieved paragraph can contain the target correction or domain term. Since the retriever is fine-tuned to retrieve target-like sentences for noisy inputs, it will retrieve this paragraph for that sentence, providing the LLM with an oracle-like hint. All LEMON results in Table 2, and the abstract's claim that RAIR outperforms current approaches in domain spelling correction, therefore rest on a confounded evaluation. The authors should rebuild the LEMON corpus from external corpora or from a training split and re-run all LEMON experiments.
  2. [Section 4.1, Table 2] The paper removes all LEMON sentence pairs whose source and target lengths differ, but it does not state whether the baseline numbers in Table 2 were recomputed on this filtered subset or copied from prior publications. If the reported ReLM, MacBERT, BERT-MFT, and GPT-3.5 scores are from the original unfiltered LEMON, the comparison is invalid because the test sets differ. The authors must recompute all baselines on the exact filtered subset used for RAIR and report the filtered statistics.
  3. [Section 4.2, Table 2] The claim of outperforming 'current approaches' is not supported by the selected baselines. The comparison omits RSpell (Song et al., 2023), a retrieval-augmented domain-adaptive CSC method, and C-LLM (Li et al., 2024), both cited in the paper, as well as other recent CSC-specific LLM approaches. Without these comparisons, the SOTA claim is not established; the authors should add the most relevant competitors or temper the claim.
  4. [Algorithm 2, Table 6] In the zero-shot setting (LEMON, ASR N-best), Algorithm 2 sets non-retrieval correction as the first-pass method and uses retrieval only as a fallback when the length check fails. The paper does not report how often retrieval is actually triggered, and the ablation in Table 6 shows that removing retrieval changes LEMON F1 by at most 1.5 points, whereas removing MLR changes it by several points. The attribution of the reported gains to 'retrieval-augmented' correction is therefore unclear; the authors should report the usage frequency of each method and discuss the contribution of each module accordingly.
minor comments (4)
  1. [Section 4.3, Table 3] The CERR column in Table 3 has inconsistent sign conventions: the definition in Section 4.3 produces negative values when the improved CER is higher, but the table reports positive percentages such as '+68.5' for GPT-3.5, which is confusing and should be corrected.
  2. [Section 4.3] The description of sentence-level precision, recall, and F1 is unclear; if a prediction is scored as correct only when it exactly matches the reference, precision and recall are identical to sentence-level accuracy and F1 is redundant. The authors should clarify the metric definition or switch to the conventional character-level metrics.
  3. [Appendix A] The Chinese prompts contain minor typographical errors, e.g., '用户输的句' in the spelling prompt should be '用户输入的句子'.
  4. [Section 4.4] The paper says all experiments are conducted on two NVIDIA 3090 GPUs, but the majority of the reported model calls are API-based (GPT-4o, GPT-3.5, Qwen, DeepSeek); the authors should clarify which components were run locally and which were run through APIs.

Circularity Check

1 steps flagged · score 6.0 of 10

The LEMON zero-shot evaluation is transductive: the retrieval corpus is generated from each test source sentence and retrieved for that same sentence, so the claimed domain-spelling gains partly reduce to self-retrieval of GPT-4o paraphrases rather than zero-shot generalization.

  1. other [Section 3.2, Creation of Retrieval Corpus (LEMON protocol)]
    "For datasets lacking training data, the generative capability of LLMs like GPT-4o is leveraged to construct background descriptions for each test source sentence. Each generated paragraph then serves as the retrieval result for its corresponding sentence, ensuring no data leakage occurs."

    On LEMON, which has no training set, the retrieval corpus is built from the test source sentences themselves. The paper makes the paragraph generated from a test sentence 'the retrieval result for its corresponding sentence,' so the retrieved context is a function of the query itself: r(x) = GPT4o(x), and the final prediction is y = LLM(x, r(x)). GPT-4o, asked for a coherent background paragraph about an erroneous source, can normalize the error while paraphrasing, so the context can contain the corrected term; the retriever, fine-tuned to retrieve correct sentences for noisy inputs, will rank this self-generated paragraph first.

full rationale

The core RAIR machinery is not circular. On ECSpell, the retrieval corpus is built from training pairs, THUOCL domain terms, and GPT-4o expansions of correct training sentences; the retriever is fine-tuned on those pairs, and MLR iteratively enforces a length reported from the input. Those components are evaluated against held-out ECSpell test data and against external ASR and CSEC benchmarks, so they do not reduce to their inputs. The circularity is localized to the LEMON evaluation in Section 3.2. Since LEMON has no training set, the paper constructs the retrieval corpus by asking GPT-4o to write a background paragraph for each test source sentence and then declares that paragraph to be the retrieval result for that sentence. The retrieved context is therefore generated from the very sentence being corrected. GPT-4o can repair or normalize the error in such a paragraph, and the fine-tuned retriever is explicitly trained to retrieve correct sentences for noisy inputs, so the paragraph is likely to be selected and to carry correction-relevant tokens. This makes the LEMON results a form of transduction or test-set leakage rather than zero-shot generalization. The paper's statement 'ensuring no data leakage occurs' is thus not supported. Because one of the two domain-spelling benchmarks that support the headline claim is compromised while the remaining evaluation is independent, the overall circularity score is 6 rather than higher.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The central claim depends on a set of hand-chosen hyperparameters (retriever negative sampling size, learning rate, temperature, epochs, top-k, MLR round limit) and on assumptions about the reliability of GPT-4o-generated retrieval content. The most fragile item is the assertion that generating background descriptions from each LEMON test sentence and then retrieving them does not leak information; this is not an artifact of the pipeline but a stated design choice. No genuinely new entities are introduced.

free parameters (7)
  • negative sampling size k (retriever) = 5
    Section 4.4 sets k=5 for retriever fine-tuning; affects quality of learned retrieval patterns.
  • retriever learning rate = 1e-5
    Section 4.4; standard hyperparameter.
  • retriever temperature tau = 0.2
    Section 4.4; controls contrastive sharpness.
  • retriever training epochs = 2
    Section 4.4.
  • retrieval top-k = 5 (implied)
    Top-k not given in main text; Table 9 uses Hit@5, and ASR selects top 5, so k=5 is implied.
  • MLR round limit m = 4 (used in analysis)
    Algorithm 1 parameter; Appendix E reports corrections within four rounds.
  • adaptive selection iteration limit k = not reported
    Algorithm 2 uses k as a threshold; no value given in the paper.
assumptions (4)
  • domain assumption Fine-tuning bge-m3 on positive/negative pairs improves retrieval robustness for noisy inputs.
    Section 3.3 relies on this; no theoretical guarantee.
  • domain assumption GPT-4o-generated domain explanations and sentence expansions are valid retrieval content.
    Section 3.2 constructs retrieval corpus via GPT-4o; quality not audited.
  • ad hoc to paper For datasets without training data, generating background descriptions from each test source sentence introduces no data leakage.
    Section 3.2 final paragraph asserts no leakage, but using test input to construct retrieval content is a leak; this is a load-bearing problem.
  • domain assumption Sentence-level exact-match F1 is an appropriate evaluation criterion.
    Section 4.3; exact match is harsh but standard.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RAIR: Retrieval-Augmented Iterative Refinement for Chinese Spelling Correction." pith.science (2026). https://pith.science/paper/CYS5HV4Y

@misc{pith2026250418938,
  author       = {Pith},
  title        = {Pith review of: RAIR: Retrieval-Augmented Iterative Refinement for Chinese Spelling Correction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CYS5HV4Y}},
  note         = {Machine review of arXiv:2504.18938}
}
read the original abstract

Chinese Spelling Correction (CSC) aims to detect and correct erroneous tokens in sentences. Traditional CSC focuses on equal length correction and uses pretrained language models (PLMs). While Large Language Models (LLMs) have shown remarkable success in identifying and rectifying potential errors, they often struggle with adapting to domain-specific corrections, especially when encountering terminologies in specialized domains. To address domain adaptation, we propose a \textbf{R}etrieval-\textbf{A}ugmented \textbf{I}terative \textbf{R}efinement (RAIR) framework. Our approach constructs a retrieval corpus adaptively from domain-specific training data and dictionaries, employing a fine-tuned retriever to ensure that the retriever catches the error correction pattern. We also extend equal-length into variable-length correction scenarios. Extensive experiments demonstrate that our framework outperforms current approaches in domain spelling correction and significantly improves the performance of LLMs in variable-length scenarios.

Figures

Figures reproduced from arXiv: 2504.18938 by the authors.

Figure 1
Figure 1. Examples of Chinese error correction in dif [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Structure diagram of RAIR, including several modules such as retrieval corpus, retriever, multi-round [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Construct positive and negative samples of [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. CSRP: Chain-of-Thought Reasoning for Chinese Text Correction via Reinforcement Learning with Efficiency-Aware Rewards

    cs.CL 2026-04 conditional novelty 6.0 of 10

    A 4B model trained with balanced CPT, CoT-SFT, and efficiency-aware GRPO reaches 50.99 F0.5 on NACGEC and 59.61 F1 on CSCD, beating larger models and GPT-4.

Reference graph

Works this paper leans on

33 extracted references · 12 canonical work pages · cited by 1 Pith paper

  1. [1]

    Tao-Hsing Chang, Hsueh-Chih Chen, and Cheng-Han Yang. 2015. https://doi.org/10.18653/v1/W15-3109 Introduction to a proofreading tool for C hinese spelling check task of SIGHAN -8 . In Proceedings of the Eighth SIGHAN Workshop on C hinese Language Processing , pages 50--55, Beijing, China. Association for Computational Linguistics

    work page doi:10.18653/v1/w15-3109 2015

  2. [2]

    Yiming Cui, Wanxiang Che, Ting Liu, Bing Qin, Shijin Wang, and Guoping Hu. 2020. https://doi.org/10.18653/v1/2020.findings-emnlp.58 Revisiting pre-trained models for C hinese natural language processing . In Findings of the Association for Computational Linguistics: EMNLP 2020, pages 657--668, Online. Association for Computational Linguistics

    doi:10.18653/v1/2020.findings-emnlp.58 2020

  3. [3]

    DeepSeek-AI, Aixin Liu, Bei Feng, Bing Xue, et al. 2025. http://arxiv.org/abs/2412.19437 Deepseek-v3 technical report

    arXiv 2025

  4. [4]

    Guanting Dong, Keming Lu, Chengpeng Li, Tingyu Xia, Bowen Yu, Chang Zhou, and Jingren Zhou. 2024 a . Self-play with execution feedback: Improving instruction-following capabilities of large language models. arXiv preprint arXiv:2406.13542

    arXiv 2024

  5. [5]

    Ming Dong, Yujing Chen, Miao Zhang, Hao Sun, and Tingting He. 2024 b . http://arxiv.org/abs/2403.08492 Rich semantic knowledge enhanced large language models for few-shot chinese spell checking

    arXiv 2024

  6. [6]

    Ming Dong, Zhiwei Cheng, Changyin Luo, and Tingting He. 2025. https://aclanthology.org/2025.coling-main.717/ Retrieval-augmented generation for large language model based few-shot C hinese spell checking . In Proceedings of the 31st International Conference on Computational Linguistics, pages 10767--10780, Abu Dhabi, UAE. Association for Computational Linguistics

    work page 2025

  7. [7]

    Zishuo Feng and Feng Cao. 2025. http://arxiv.org/abs/2411.11770 Cnmbert: A model for hanyu pinyin abbreviation to character conversion task

    arXiv 2025

  8. [8]

    Yuzhong Hong, Xianguo Yu, Neng He, Nan Liu, and Junhui Liu. 2019. https://doi.org/10.18653/v1/D19-5522 FASPell : A fast, adaptable, simple, powerful chinese spell checker based on DAE -decoder paradigm . In Proceedings of the 5th Workshop on Noisy User-generated Text (W- NUT 2019) , pages 160--169. Association for Computational Linguistics

    work page doi:10.18653/v1/d19-5522 2019

Show all 33 references
  1. [9]

    Yong Hu, Fandong Meng, and Jie Zhou. 2022. Cscd-ime: correcting spelling errors generated by pinyin ime. arXiv preprint arXiv:2211.08788

    2022 arXiv

  2. [10]

    Kunting Li, Yong Hu, Liang He, Fandong Meng, and Jie Zhou. 2024. http://arxiv.org/abs/2406.16536 C-llm: Learn to check chinese spelling errors character by character

    2024 arXiv

  3. [11]

    Yinghui Li, Haojing Huang, Shirong Ma, Yong Jiang, Yangning Li, Feng Zhou, Hai-Tao Zheng, and Qingyu Zhou. 2023. http://arxiv.org/abs/2307.09007 On the (in)effectiveness of large language models for chinese text correction

    2023 arXiv

  4. [12]

    Junhong Liang. 2024. http://arxiv.org/abs/2412.03230 Perl: Pinyin enhanced rephrasing language model for chinese asr n-best error correction

    2024 arXiv

  5. [13]

    Junhong Liang, Zhu Junnan, Feifei Zhai, Nanchang Cheng, Chengqing Zong, and Yu Zhou. 2024. https://doi.org/10.3233/FAIA240952 A Hybrid Approach towards Chinese Spelling and Splitting Error Correction , pages 3883--3890. ECAI

    2024 doi

  6. [14]

    Qi Lv, Ziqiang Cao, Lei Geng, Chunhui Ai, Xu Yan, and Guohong Fu. 2023. https://doi.org/10.1145/3564271 General and domain-adaptive chinese spelling check with error-consistent pretraining . ACM Trans. Asian Low-Resour. Lang. Inf. Process., 22(5)

    2023 doi

  7. [15]

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. J. Mach. Learn. Res., 21(1)

    2020

  8. [16]

    Siqi Song, Qi Lv, Lei Geng, Ziqiang Cao, and Guohong Fu. 2023. Rspell: Retrieval-augmented framework for domain adaptive chinese spelling check. In Natural Language Processing and Chinese Computing, pages 551--562, Cham. Springer Nature Switzerland

    2023

  9. [17]

    Zijun Sun, Xiaoya Li, Xiaofei Sun, Yuxian Meng, Xiang Ao, Qing He, Fei Wu, and Jiwei Li. 2021. https://doi.org/10.18653/v1/2021.acl-long.161 C hinese BERT : C hinese pretraining enhanced by glyph and P inyin information . In Proceedings of the 59th Annual Meeting of the Associ...

    2021 doi

  10. [18]

    Zhiyuan Tang, Dong Wang, Shen Huang, and Shidong Shang. 2024. https://doi.org/10.21437/Interspeech.2024-987 Pinyin regularization in error correction for chinese speech recognition with large language models . In Interspeech 2024, pages 1910--1914. ISCA

    2024 doi

  11. [19]

    Qwen Team, An Yang, Baosong Yang, Beichen Zhang, et al. 2025. http://arxiv.org/abs/2412.15115 Qwen2.5 technical report

    2025 arXiv

  12. [20]

    Yuen-Hsien Tseng, Lung-Hao Lee, Li-Ping Chang, and Hsin-Hsi Chen. 2015. https://doi.org/10.18653/v1/W15-3106 Introduction to SIGHAN 2015 bake-off for C hinese spelling check . In Proceedings of the Eighth SIGHAN Workshop on C hinese Language Processing , pages 32--37, Beijing,...

    2015 doi

  13. [21]

    Hongqiu Wu, Shaohua Zhang, Yuchen Zhang, and Hai Zhao. 2023. http://arxiv.org/abs/2305.17721 Rethinking masked language modeling for chinese spelling correction

    2023 arXiv

  14. [22]

    Shih-Hung Wu, Chao-Lin Liu, and Lung-Hao Lee. 2013. https://aclanthology.org/W13-4406 C hinese spelling check evaluation at SIGHAN bake-off 2013 . In Proceedings of the Seventh SIGHAN Workshop on C hinese Language Processing , pages 35--42, Nagoya, Japan. Asian Federation of N...

    2013

  15. [23]

    Heng-Da Xu, Zhongli Li, Qingyu Zhou, Chao Li, Zizhen Wang, Yunbo Cao, Heyan Huang, and Xian-Ling Mao. 2021. http://arxiv.org/abs/2105.12306 [cs] Read, listen, and see: Leveraging multimodal information helps chinese spell checking

    2021 arXiv

  16. [24]

    Zhu Xu, Zhiqiang Zhao, Zihan Zhang, Yuchi Liu, Quanwei Shen, Fei Liu, Yu Kuang, Jian He, and Conglin Liu. 2024. http://arxiv.org/abs/2411.17679 Enhancing character-level understanding in llms through token internal structure learning

    2024 arXiv

  17. [25]

    Xunjian Yin, Xinyu Hu, Jin Jiang, and Xiaojun Wan. 2024. http://arxiv.org/abs/2211.07843 Error-robust retrieval for chinese spelling check

    2024 arXiv

  18. [26]

    Liang-Chih Yu, Lung-Hao Lee, Yuen-Hsien Tseng, and Hsin-Hsi Chen. 2014. https://doi.org/10.3115/v1/W14-6820 Overview of SIGHAN 2014 bake-off for C hinese spelling check . In Proceedings of the Third CIPS - SIGHAN Joint Conference on C hinese Language Processing , pages 126--13...

    2014 doi

  19. [27]

    Yang Yuhang, Peng Yizhou, Eng Siong Chng, and Xionghu Zhong. 2024. http://arxiv.org/abs/2409.16005 Bridging speech and text: Enhancing asr with pinyin-to-character pre-training in llms

    2024 arXiv

  20. [28]

    Shaohua Zhang, Haoran Huang, Jicong Liu, and Hang Li. 2020. http://arxiv.org/abs/2005.07421 Spelling error correction with soft-masked bert

    2020 arXiv

  21. [29]

    Xiang Zhang, Junbo Zhao, and Yann LeCun. 2015. http://arxiv.org/abs/1509.01626 Character-level convolutional networks for text classification

    2015 arXiv

  22. [30]

    Houquan Zhou, Zhenghua Li, Bo Zhang, Chen Li, Shaopeng Lai, Ji Zhang, Fei Huang, and Min Zhang. 2024. http://arxiv.org/abs/2410.04027 A simple yet effective training-free prompt-free approach to chinese spelling correction based on large language models

    2024 arXiv

  23. [31]

    Chenxi Zhu, Ziqiang Ying, Boyu Zhang, and Feng Mao. 2022. https://doi.org/10.18653/v1/2022.findings-acl.98 MDCS pell: A multi-task detector-corrector framework for C hinese spelling correction . In Findings of the Association for Computational Linguistics: ACL 2022, pages 1244...

    2022 doi

  24. [32]

    URL: " 'urlintro :=

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

  25. [33]

    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 16, 2026 · model on record in the stance chip above.