Data Augmentation for Spoken Grammatical Error Correction

Kate M. Knill; Mark J.F. Gales; Mengjie Qian; Penny Karanasou; Stefano Bann\`o

REVIEW 2 major objections 5 minor 36 references

Data Augmentation for Spoken Grammatical Error Correction

T0 review · 2 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A fully automatic pipeline generates audio-text pairs with learner-like errors and disfluencies, and those synthetic pairs improve spoken grammatical error correction in cascaded, semi-cascaded, and end-to-end systems.

desk verdict Useful automated augmentation pipeline for low-resource spoken GEC, worth peer review; the headline end-to-end gain is confounded by fine-tuning and the clean comparison is small. read the letter →

arxiv 2507.19374 v1 pith:NTZUHBFW submitted 2025-07-25 cs.CL cs.AIcs.SDeess.AS

Penny Karanasou , Mengjie Qian , Stefano Bannò , Mark J.F. Gales , Kate M. Knill This is my paper · ORCID

classification cs.CLcs.AIcs.SDeess.AS

keywords spokengrammaticalerrorcorrectiondataaugmentationreversetext-to-speechdisfluencyspeakersimilarityfilteringobjectiveevaluationmetricssecondlanguageassessment

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

Spoken grammatical error correction (SGEC) is under-resourced compared with text GEC, so this paper tries to manufacture the missing training data automatically. Its proposed pipeline first runs a text-to-error 'reverse GEC' model over clean transcriptions, then re-inserts disfluencies at positions taken from real learner speech, then synthesizes audio with a voice-cloned text-to-speech system. The paper also proposes four objective metrics to choose among candidate generated datasets without human listening. On the Speak & Improve Corpus 2025, the augmented text and audio improve cascaded, semi-cascaded, and end-to-end SGEC models, with the best end-to-end model reaching F0.5 41.93 versus 41.39 for the original-data baseline. If this holds, SGEC systems can be improved without costly manual annotation and without shifting assessment scores.

What carries the argument

The central mechanism is a reverse GEC model: a BART-large sequence-to-sequence model trained with input and output swapped so that clean text becomes errorful text with the same error-type distribution as the training corpus. Around this core, the pipeline inserts disfluencies at positions taken from the original aligned data, then uses F5-TTS, a flow-matching non-autoregressive TTS with zero-shot voice cloning, to synthesize audio that preserves the original speaker's voice and accent. The four proposed metrics—speaker-embedding cosine distance, ASR word error rate, a BERT-based text grader, and a wav2vec2-based audio grader—are the selection machinery that decides whether a candidate generated corpus is usable, and they identify F5-TTS as the best among the systems compared.

What would settle it

Train a GEC system only on the augmented corpus and a matched system only on original annotated learner speech, then test both on a held-out set of real learner utterances containing error types not seen in the generation distribution; a clear gap would show the generated errors are not representative. A simpler check is a human Turing-style rating of whether generated errorful audio-text pairs come from real learners.

Watch

Extended reading notes

Core claim

The central claim, stated as a fair reader would state it: learner-like training data for SGEC can be synthesized rather than collected. Reversing a BART-large GEC model produces errorful text whose error-type distribution resembles the original corpus, copying disfluency positions preserves spoken disfluency patterns, and F5-TTS voice cloning preserves speaker identity; the resulting audio-text pairs are close enough to original data that they improve all three SGEC architectures tested. The decisive numbers are in Tables 3 and 4: written augmentation raises F0.5 on manual transcriptions from 52.01 to 54.94 and in every cascaded and semi-cascaded ASR condition, while spoken augmentation with cosine-distance filtering and fine-tuning pushes end-to-end F0.5 from 41.39 to 41.93. The objective metrics corroborate the choice of F5-TTS, which shows the smallest speaker-embedding distance and the closest WER to original audio among the TTS systems compared.

Load-bearing premise

The invented errors must be the same kind a real second-language learner would make; the paper checks this only with a corrector trained on the same data as the error generator, so the check could pass even if the generated errors are unnatural.

Editorial extensions

If this is right

Adding reverse-GEC text to the training set improves written GEC on the eval set, raising F0.5 from 52.01 to 54.94 on manual transcriptions.
In cascaded and semi-cascaded SGEC pipelines, augmented text helps across ASR configurations: Whisper-small with disfluency detection rises from F0.5 22.11 to 28.28, Whisper-large from 24.37 to 30.31, and the fluent-tuned Whisper from 35.89 to 38.72.
For end-to-end SGEC, generated F5-TTS audio alone roughly matches the original-data baseline, but filtering by speaker similarity lifts F0.5 to 41.65, and fine-tuning on original data after augmentation reaches 41.93.
The four objective metrics can rank candidate generated datasets without subjective listening, and the paper's chosen F5-TTS system is the one that scores best on them.
Because the augmented corpus preserves SLA grader scores, it can enrich learner speech data without altering the distribution of assessment outcomes.

Reading between the lines

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

A stronger validation of the reverse-GEC errors is available than the paper's self-consistency check: have annotators or a GEC model trained on independent data judge whether the generated errors look like learner errors, since the current checker shares training data with the generator.
The speaker-similarity filter is used as a binary threshold; one testable extension is to weight generated samples by the continuous cosine distance instead, which may preserve more data and improve end-to-end training.
Because the pipeline separates error generation, disfluency placement, and synthesis, each module could be evaluated in isolation, for example by learning disfluency insertion from the corpus rather than copying positions, to see whether more varied disfluencies improve robustness.

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

Quick take: the reverse-GEC + disfluency insertion + F5-TTS pipeline is a sensible, genuinely useful recipe for a task that lacks training data, and the four objective metrics (speaker similarity AUC, ASR WER, BERT/wav2vec SLA score correlation) are a nice addition that goes beyond anecdotal listening tests. The paper is the first to apply this combination to the S&I Corpus, and it shows consistent improvements across cascaded, semi-cascaded, and end-to-end SGEC setups. Written GEC gains in Table 3 are more convincing than the spoken gains.

The real soft spot is Table 4. Row 4 (F0.5 41.93) mixes generated-audio pretraining with fine-tuning on the original training set, so it cannot isolate augmentation. The stress-test note is half right: row 3 (41.65) is a clean augmentation comparison and shows a small +0.26 gain over the 41.39 baseline, so the claim isn't empty. But that gain is modest, there are no error bars or significance tests across any of the runs, and the unfiltered condition (row 2) is a wash. The paper would be stronger with a fully crossed design: with and without augmentation, with and without fine-tuning, and repeated seeds.

The reverse-GEC sanity check in Table 1 is circular in the sense that the checker and the generator are trained on the same BEA-2019 data family; it shows internal consistency, not that generated errors match real learner errors. That's worth a sentence of caution but not fatal, because the downstream task evaluation is held-out. Also minor: Section 3 says the metrics are computed on a dev subset of "~3300K utterances" — that's presumably a typo (3.3M is far larger than the corpus), and it should be corrected.

The paper could be more open: no code, no data, no released filters. That limits reproducibility, especially for a pipeline that depends on many small choices (checkpoint selection, similarity threshold cos04, TTS prompt handling).

Overall: a solid, incremental contribution to a genuinely under-resourced task. A serious referee should engage with it. My recommendation: send to peer review; expect revision with a proper ablation, significance tests, and a fix for the typo.

Referee Report

2 major / 5 minor

Summary. The paper proposes a fully automated data augmentation pipeline for spoken grammatical error correction (SGEC): a reverse GEC model generates erroneous text from clean transcriptions, a disfluency insertion module adds hesitations, repetitions, false starts and incomplete sentences at original positions, and F5-TTS synthesizes audio with voice cloning. The authors introduce four objective metrics (speaker-verification cosine distance, ASR WER, BERT-based text scoring, and wav2vec2 audio scoring) to compare candidate generated datasets. They evaluate the augmented data on written GEC and on cascaded, semi-cascaded, and end-to-end SGEC systems using the Speak & Improve Corpus 2025. The reported results show consistent gains for text-only augmentation in cascaded and semi-cascaded pipelines, while the end-to-end gains in Table 4 are small and the best result confounds augmentation with fine-tuning on the original data.

Significance. If the result holds, the pipeline offers a valuable answer to the data scarcity problem in SGEC by reusing strong text GEC resources, TTS with voice cloning, and a public L2 speech corpus. The four objective metrics are a useful contribution in themselves, as they give a reusable, largely automated way to select between generated spoken datasets. The paper is also reasonably reproducible: the components (BART, BEA-2019, Whisper, F5-TTS, the S&I Corpus) are public, and the metric computations are described concretely. The main unresolved point is experimental control: the end-to-end SGEC claim rests on a comparison that mixes two interventions, and the only non-confounded comparison in Table 4 does not show a gain. The reverse-GEC quality check is also self-referential, so the representativeness of the generated errors remains an open risk.

major comments (2)

[Section 4.2, Table 4] The paper's central end-to-end SGEC claim is confounded. Row 4 (F0.5 41.93) is obtained by taking the model trained on Train+Dev+Gen(cos04) and fine-tuning it on Train+Dev, while the baseline row 1 (41.39) is trained on Train+Dev without an equivalent fine-tuning stage. The only non-confounded augmentation comparison, row 2 versus row 1, shows a slight degradation (41.34 vs 41.39). Therefore the sentence "The results show that using augmented data improves performance" in Section 4.2 is not supported by Table 4 for end-to-end models. Please add a no-augmentation control that is fine-tuned on Train+Dev with the same two-stage schedule, and report variance or significance across seeds; the current +0.54 difference is small and could be noise.
[Section 2.1, Table 1] The sanity check that reverse GEC outputs can be corrected by the S&I 2025 GEC system is circular. Both the reverse GEC model and the GEC checker are trained on the same BEA-2019 data, so a high ERRANT F0.5 on the reverse-GEC outputs demonstrates self-consistency of the model family rather than naturalness of the generated errors. Because the usefulness of the augmented data for real L2 speech depends on the synthetic errors resembling genuine learner errors, please provide an external validation, for example human annotation of a sample of generated utterances or a comparison of error-type distributions against a held-out S&I eval subset that was not used to train the checker.

minor comments (5)

[Section 3] The text says the metrics are computed on a subset of the S&I dev set of "∼ 3300K utterances"; 3.3 million utterances is implausible for a dev subset and this is likely a typo (perhaps 3.3K or 330K). Please clarify, since the size affects the reliability of the metric comparisons in Figures 3 and 4.
[Section 2.2] "We hypotheses that minimizing text edits may ease voice cloning" should read "We hypothesize that."
[Section 3.3] Figure 4 is described as showing "strong positive correlations," but no correlation coefficients are reported; please include numeric values such as Pearson r for both the BERT and wav2vec2 scatter plots.
[Section 3.2, Table 2] F5-TTS shows a much higher insertion rate (8.7) than the original (1.4), attributed to prompt words at the start of generated audio. Please clarify whether these prompt words are removed from the generated audio and text before the data are used for SGEC training, since otherwise they would be a systematic artifact in the augmented training set.
[Section 2.3] The comparison of TTS systems is described only as "informally evaluated through subjective listening"; please provide details on the number of listeners, the stimuli, and the criteria so that this comparison can be assessed and reproduced.

Circularity Check

1 steps flagged · score 2.0 of 10

Reverse-GEC sanity check is a self-consistency loop; the main SGEC claim rests on held-out S&I evaluation and is not circular.

self definitional [Section 2.1, final paragraph (Table 1 discussion)]
"As an extra sanity check, the generated transcriptions from the reverse GEC are passed through the GEC system trained for the S&I 2025 Challenge. This is to verify that the generated errors can be corrected by the existing GEC system, indicating that they align with the error types seen in its training set. Table 1 reports the F 0.5 score calculated based on the ERRANT edits [19]. Interestingly, the score on the generated transcriptions outperforms the original transcriptions, giving a strong indication that the generated errors are of the expected types."

The reverse GEC module is a BART-large model previously used for GEC in the S&I 2025 Challenge, with input and output reversed, and is trained on the same ~1M BEA-2019 utterances used for the GEC model. The sanity check then asks whether the S&I 2025 GEC system can correct the generated errors. Because both models are trained on the same error-correction distribution, correctability is expected by construction: the reverse model maps corrected text back into the same error patterns the forward model was trained to repair. The check therefore establishes internal consistency between two models trained on identical data, not that the generated errors match real L2 learner errors.

full rationale

The paper's central claim, that the proposed augmentation pipeline improves written GEC and spoken GEC, is evaluated on the held-out S&I eval set in Tables 3 and 4. This external evaluation is not circular: the augmented-data models are compared against baselines on real L2 learner speech. The one internal loop is the Section 2.1 'extra sanity check', where reverse-GEC output is validated by passing it through the S&I 2025 GEC system. Since the reverse model and the GEC checker are trained on the same BEA-2019 data, high correctability is a self-consistency result rather than independent evidence of naturalness or learner representativeness. This raises the circularity burden modestly, but it is not load-bearing for the main SGEC result, which is tested on held-out data. The other objective metrics (ASR WER, BERT and wav2vec SLA graders) use Linguaskill-trained models or held-out comparisons, so they do not introduce a circular loop. The Table 4 best row also confounds augmentation with fine-tuning on the original training set, but that is a missing-control issue rather than circularity and should be addressed with a no-augmentation fine-tuned control. Overall, the only notable circular step is the reverse-GEC self-check, giving a score of 2.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The central results depend on the quality of the reverse GEC model and the TTS voice cloning, both of which are pretrained off-the-shelf components used without fine-tuning. The only tuned parameter is the cosine threshold for filtering, which is set based on dev-set speaker similarity. No new entities are postulated.

free parameters (2)

speaker similarity filtering threshold (cos04) = 0.04
Chosen on the S&I dev set from the speaker verification analysis in Section 3.1; used to filter generated audio-text pairs in Table 4.
reverse GEC checkpoint selection criterion = >80% new errors
A checkpoint is selected that produces new errors in over 80% of generated utterances (Section 2.1); this is a model-selection criterion based on the desired output property.

assumptions (3)

domain assumption The S&I Corpus manual transcriptions, disfluency tags, and grammar error annotations are accurate.
The pipeline uses these annotations as ground truth for training and evaluation (Section 4.1).
domain assumption ASR errors from Whisper dsf are consistent with the baseline, so WER increases are attributed to the TTS system.
Stated in Section 3.2; if ASR errors vary by condition, the WER comparison is invalid.
domain assumption The BERT and wav2vec2 SLA graders trained on Linguaskill data provide valid proficiency scores for both original and generated audio.
Used as objective metrics in Section 3.3; their validity for synthetic TTS audio is assumed.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Data Augmentation for Spoken Grammatical Error Correction." pith.science (2026). https://pith.science/paper/NTZUHBFW

@misc{pith2026250719374,
  author       = {Pith},
  title        = {Pith review of: Data Augmentation for Spoken Grammatical Error Correction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NTZUHBFW}},
  note         = {Machine review of arXiv:2507.19374}
}

read the original abstract

While there exist strong benchmark datasets for grammatical error correction (GEC), high-quality annotated spoken datasets for Spoken GEC (SGEC) are still under-resourced. In this paper, we propose a fully automated method to generate audio-text pairs with grammatical errors and disfluencies. Moreover, we propose a series of objective metrics that can be used to evaluate the generated data and choose the more suitable dataset for SGEC. The goal is to generate an augmented dataset that maintains the textual and acoustic characteristics of the original data while providing new types of errors. This augmented dataset should augment and enrich the original corpus without altering the language assessment scores of the second language (L2) learners. We evaluate the use of the augmented corpus both for written GEC (the text part) and for SGEC (the audio-text pairs). Our experiments are conducted on the S\&I Corpus, the first publicly available speech dataset with grammar error annotations.

Figures

Figures reproduced from arXiv: 2507.19374 by the authors.

**Figure 1.** Cascaded spoken GEC process (top) and automatic spoken GEC data generation process (bottom). a GEC module to correct the grammar. [6, 7] leverage largely available audio recordings to generate pseudo GEC transcriptions using a semi-cascaded pipeline with Whisper and a textbased GEC model. In this paper, a reverse pipeline is proposed to automatically generate data by introducing grammatical errors and disfluencie… view at source ↗

**Figure 2.** Distribution plot of error categories in ref (original transcriptions) and hyp (transcriptions with generated errors) As an extra sanity check, the generated transcriptions from the reverse GEC are passed through the GEC system trained for the S&I 2025 Challenge. This is to verify that the generated errors can be corrected by the existing GEC system, indicating that they align with the error types seen in its traini… view at source ↗

**Figure 3.** presents the cumulative plots of cosine distances for the systems under comparison. The F5-TTS curve is the steepest on the left side, indicating a higher proportion of closely matched pairs between original and generated data. This observation is further supported by the Area Under the Curve (AUC), where F5-TTS yields the lowest value. A lower AUC value indicates that the distance values are clustered towards th… view at source ↗

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 24 canonical work pages

[1]

Introduction Automatic spoken language assessment (SLA) is the task of grading second language (L2) learners and providing them feed- back in an automatic way, without human expert knowledge. Grammatical error correction (GEC) is an important part of SLA and a well-established research area [1], supported by a number of shared tasks such as CoNLL-2014 [2]...

work page Pith review arXiv 2014
[2]

Each module is detailed below

Data Augmentation Pipeline We present a model-based data augmentation pipeline that gen- erates aligned text and audio data with added grammatical er- rors and disfluencies. Each module is detailed below. 2.1. Reverse GEC Module The first step adopts a reverse GEC module to automati- cally generate text with grammatical errors. A BART-large model [17], pr...

work page 2025
[3]

These metrics offer a systematic approach to selecting the optimal set of spoken augmented data from multiple candidates

Objective Metrics for Spoken Augmented Data Analysis This section proposes four metrics to evaluate and compare dif- ferent versions of the generated spoken augmented data. These metrics offer a systematic approach to selecting the optimal set of spoken augmented data from multiple candidates. In this study, we will only change the TTS model used in the d...

work page
[4]

Spoken GEC evaluation 4.1. Corpus The Speak & Improve (S&I) Corpus 2025 is a comprehensive, public dataset of second-language (L2) learner speech, designed to facilitate research in automated spoken language assessment and feedback [16]. Collected through the Speak & Improve platform between 2019 and 2024 [30], the corpus features di- verse learner audio ...

work page 2025
[5]

This pipeline consists of a reverse GEC model, a disfluency addition module and a TTS component

Conclusion In this work, we propose an automatic pipeline to generate audio-text pairs for training end-to-end SGEC models. This pipeline consists of a reverse GEC model, a disfluency addition module and a TTS component. Four metrics are then proposed to evaluate the quality of spoken augmented data and to select the optimal set of generated data. Various...

work page
[6]

Grammatical Error Correction: A Survey of the State of the Art,

C. Bryant, Z. Yuan, M. R. Qorib, H. Cao, H. T. Ng, and T. Briscoe, “Grammatical Error Correction: A Survey of the State of the Art,” Computational Linguistics, vol. 49, no. 3, pp. 643–701, 09 2023

work page 2023
[7]

The CoNLL-2014 Shared Task on Grammatical Error Correction,

H. T. Ng, S. M. Wu, T. Briscoe, C. Hadiwinoto, R. H. Susanto, and C. Bryant, “The CoNLL-2014 Shared Task on Grammatical Error Correction,” in Proceedings of the 18th conference on computational natural language learning: shared task, 2014, pp. 1–14. [Online]. Available: https://aclanthology.org/W14-1701/

work page 2014
[8]

The BEA- 2019 Shared Task on Grammatical Error Correction,

C. Bryant, M. Felice, Ø. E. Andersen, and T. Briscoe, “The BEA- 2019 Shared Task on Grammatical Error Correction,” inProceed- ings of the fourteenth workshop on innovative use of NLP for building educational applications, 2019, pp. 52–75

work page 2019

Show all 36 references

[9]

Shared task on Multilingual Grammatical Error Correc- tion 2025,

E. V olodina, A. Masciolini, A. Caines et al. , “Shared task on Multilingual Grammatical Error Correc- tion 2025,” https://www.aclweb.org/portal/content/ shared-task-multilingual-grammatical-error-correction-2025, 2025

2025
[10]

Spoken Language ‘Gram- matical Error Correction’,

Y . Lu, M. J. F. Gales, and Y . Wang, “Spoken Language ‘Gram- matical Error Correction’,” in Proc. INTERSPEECH 2020, 2020, pp. 3840–3844

2020
[11]

Scaling and Prompting for Improved End-to-End Spoken Grammatical Er- ror Correction,

M. Qian, R. Ma, S. Bann `o, K. M. Knill, and M. J. Gales, “Scaling and Prompting for Improved End-to-End Spoken Grammatical Er- ror Correction,” in Proc. INTERSPEECH 2025, 2025

2025
[12]

End- to-End Spoken Grammatical Error Correction,

M. Qian, R. Ma, S. Bann `o, M. J. Gales, and K. M. Knill, “End- to-End Spoken Grammatical Error Correction,” arXiv preprint arXiv:2506.18532, 2025

2025 arXiv
[13]

Minimally- Augmented Grammatical Error Correction,

R. Grundkiewicz and M. Junczys-Dowmunt, “Minimally- Augmented Grammatical Error Correction,” in Proceedings of the 5th Workshop on Noisy User-generated Text (W-NUT 2019), W. Xu, A. Ritter, T. Baldwin, and A. Rahimi, Eds. Hong Kong, China: Association for Computational Linguist...

2019
[14]

Optimizing the impact of data augmentation for low-resource grammatical error correction,

A. Solyman, M. Zappatore, W. Zhenyu, Z. Mahmoud, A. Alfatemi, A. O. Ibrahim, and L. A. Gabralla, “Optimizing the impact of data augmentation for low-resource grammatical error correction,” J. King Saud Univ. Comput. Inf. Sci. , vol. 35, no. 6, Jun. 2023. [Online]. Available: h...

2023
[15]

Comparison of Grammatical Error Correction Using Back- Translation Models,

A. Koyama, K. Hotate, M. Kaneko, and M. Komachi, “Comparison of Grammatical Error Correction Using Back- Translation Models,” in Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Student Research Workshop . Onlin...

2021
[16]

To Err Is Human, but Llamas Can Learn It Too,

A. Luhtaru, T. Purason, M. Vainikko, M. Del, and M. Fishel, “To Err Is Human, but Llamas Can Learn It Too,” in Findings of the Association for Computational Linguistics: EMNLP 2024 . Mi- ami, Florida, USA: Association for Computational Linguistics, Nov. 2024, pp. 12 466–12 481

2024
[17]

Artificial disfluency detection, uh no, disfluency generation for the masses,

T. Passali, T. Mavropoulos, G. Tsoumakas, G. Meditskos, and S. Vrochidis, “Artificial disfluency detection, uh no, disfluency generation for the masses,” Computer Speech and Language , vol. 89, p. 101711, 2025

2025
[18]

Automatic Assessment of English as a Second Lan- guage,

S. Dubov, “Automatic Assessment of English as a Second Lan- guage,” Master’s thesis, University of Cambridge, Cambridge, UK, 2021

2021
[19]

Text Generation with Speech Synthesis for ASR Data Augmentation,

Z. Huang, G. Keren, Z. Jiang, S. Jain, D. Goss-Grubbs, N. Cheng, F. Abtahi, D. Le, D. Zhang, A. D’Avirro, E. Campbell-Taylor, J. Salas, I.-E. Veliche, and X. Chen, “Text Generation with Speech Synthesis for ASR Data Augmentation,” 2023. [Online]. Available: https://arxiv.org/a...

2023 arXiv
[20]

SynthASR: Un- locking synthetic data for speech recognition,

A. Fazel, W. Yang, Y . Liu, R. Barra-Chicote, Y . Meng, R. Maas, and J. Droppo, “SynthASR: Un- locking synthetic data for speech recognition,” 2021. [Online]. Available: https://www.amazon.science/publications/ synthasr-unlocking-synthetic-data-for-speech-recognition

2021
[21]

Speak & Improve Corpus 2025: an L2 English Speech Corpus for Language Assessment and Feedback,

K. Knill, D. Nicholls, M. J. Gales, M. Qian, and P. Stroinski, “Speak & Improve Corpus 2025: an L2 English Speech Corpus for Language Assessment and Feedback,” 2024. [Online]. Available: https://doi.org/10.17863/CAM.114333

2025 doi
[22]

BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Gener- ation, Translation, and Comprehension,

M. Lewis, Y . Liu, N. Goyal, M. Ghazvininejad, A. Mohamed, O. Levy, V . Stoyanov, and L. Zettlemoyer, “BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Gener- ation, Translation, and Comprehension,” in Proceedings of the 58th Annual Meeting of the Associa...

2020
[23]

Speak & Improve Challenge 2025: Tasks and Baseline Systems,

M. Qian, K. Knill, S. Banno, S. Tang, P. Karanasou, M. J. Gales, and D. Nicholls, “Speak & Improve Challenge 2025: Tasks and Baseline Systems,” 2024. [Online]. Available: https://arxiv.org/abs/2412.11985

2025 arXiv
[24]

Automatic Annotation and Evaluation of Error Types for Grammatical Error Correction,

C. Bryant, M. Felice, and T. Briscoe, “Automatic Annotation and Evaluation of Error Types for Grammatical Error Correction,” in Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), R. Barzilay and M.-Y . Kan, Eds. Van...

2017
[25]

F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching,

Y . Chen, Z. Niu, Z. Ma, K. Deng, C. Wang, J. Zhao, K. Yu, and X. Chen, “F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching,” 2024. [Online]. Available: https://arxiv.org/abs/2410.06885

2024 arXiv
[26]

Scalable diffusion models with transform- ers,

W. Peebles and S. Xie, “Scalable diffusion models with transform- ers,” inProceedings of the IEEE/CVF international conference on computer vision, 2023, pp. 4195–4205

2023
[27]

V oiceCraft: Zero-Shot Speech Editing and Text-to-Speech in the Wild,

P. Peng, P.-Y . Huang, S.-W. Li, A. Mohamed, and D. Harwath, “V oiceCraft: Zero-Shot Speech Editing and Text-to-Speech in the Wild,” in Proceedings of the 62nd Annual Meeting of the Asso- ciation for Computational Linguistics (Volume 1: Long Papers) , 2024, pp. 12 442–12 462

2024
[28]

Speak Foreign Languages with Your Own V oice: Cross-Lingual Neural Codec Language Modeling,

Z. Zhang, L. Zhou, C. Wang, S. Chen, Y . Wu, S. Liu, Z. Chen, Y . Liu, H. Wang, J. Li, L. He, S. Zhao, and F. Wei, “Speak Foreign Languages with Your Own V oice: Cross-Lingual Neural Codec Language Modeling,” 2023. [Online]. Available: https://arxiv.org/abs/2303.03926

2023 arXiv
[29]

YourTTS: Towards zero-shot multi-speaker TTS and zero-shot voice conversion for everyone,

E. Casanova, J. Weber, C. D. Shulby, A. C. Junior, E. G ¨olge, and M. A. Ponti, “YourTTS: Towards zero-shot multi-speaker TTS and zero-shot voice conversion for everyone,” in Proceedings of the 39th International Conference on Machine Learning , ser. Proceedings of Machine Lea...

2022
[30]

XTTS: a Massively Multilingual Zero-Shot Text-to-Speech Model,

E. Casanova, K. Davis, E. G ¨olge, G. G ¨oknar, I. Gulea, L. Hart, A. Aljafari, J. Meyer, R. Morais, S. Olayemi, and J. We- ber, “XTTS: a Massively Multilingual Zero-Shot Text-to-Speech Model,” in Interspeech 2024, 2024, pp. 4978–4982

2024
[31]

Towards Human-Sounding TTS,

T. C. L. Team, “Towards Human-Sounding TTS,” https:// canopylabs.ai/model-releases, 2025, accessed: March 29, 2025

2025
[32]

pyannote.audio 2.1 speaker diarization pipeline: prin- ciple, benchmark, and recipe,

H. Bredin, “pyannote.audio 2.1 speaker diarization pipeline: prin- ciple, benchmark, and recipe,” in Interspeech 2023 , 2023, pp. 1983–1987

2023
[33]

Towards End-to-End Spoken Grammatical Error Correction,

S. Bann `o, R. Ma, M. Qian, K. M. Knill, and M. J. Gales, “Towards End-to-End Spoken Grammatical Error Correction,” in IEEE In- ternational Conference on Acoustics, Speech and Signal Process- ing (ICASSP). IEEE, 2024, pp. 10 791–10 795

2024
[34]

Ludlow, Official Quick Guide to Linguaskill

K. Ludlow, Official Quick Guide to Linguaskill . Cambridge: Cambridge University Press & Assessment, 2020

2020
[35]

Speak & Improve: L2 English Speaking Practice Tool,

D. Nicholls, K. M. Knill, M. J. F. Gales, A. Ragni, and P. Ricketts, “Speak & Improve: L2 English Speaking Practice Tool,” in Proc. INTERSPEECH 2023, 2023, pp. 3669–3670

2023
[36]

Cambridge: Cambridge University Press, 2001

Council of Europe, Common European Framework of Reference for Languages: Learning, Teaching, Assessment . Cambridge: Cambridge University Press, 2001. [Online]. Available: https: //rm.coe.int/1680459f97

2001

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [1]

Introduction Automatic spoken language assessment (SLA) is the task of grading second language (L2) learners and providing them feed- back in an automatic way, without human expert knowledge. Grammatical error correction (GEC) is an important part of SLA and a well-established research area [1], supported by a number of shared tasks such as CoNLL-2014 [2]...

work page Pith review arXiv 2014

[2] [2]

Each module is detailed below

Data Augmentation Pipeline We present a model-based data augmentation pipeline that gen- erates aligned text and audio data with added grammatical er- rors and disfluencies. Each module is detailed below. 2.1. Reverse GEC Module The first step adopts a reverse GEC module to automati- cally generate text with grammatical errors. A BART-large model [17], pr...

work page 2025

[3] [3]

These metrics offer a systematic approach to selecting the optimal set of spoken augmented data from multiple candidates

Objective Metrics for Spoken Augmented Data Analysis This section proposes four metrics to evaluate and compare dif- ferent versions of the generated spoken augmented data. These metrics offer a systematic approach to selecting the optimal set of spoken augmented data from multiple candidates. In this study, we will only change the TTS model used in the d...

work page

[4] [4]

Spoken GEC evaluation 4.1. Corpus The Speak & Improve (S&I) Corpus 2025 is a comprehensive, public dataset of second-language (L2) learner speech, designed to facilitate research in automated spoken language assessment and feedback [16]. Collected through the Speak & Improve platform between 2019 and 2024 [30], the corpus features di- verse learner audio ...

work page 2025

[5] [5]

This pipeline consists of a reverse GEC model, a disfluency addition module and a TTS component

Conclusion In this work, we propose an automatic pipeline to generate audio-text pairs for training end-to-end SGEC models. This pipeline consists of a reverse GEC model, a disfluency addition module and a TTS component. Four metrics are then proposed to evaluate the quality of spoken augmented data and to select the optimal set of generated data. Various...

work page

[6] [6]

Grammatical Error Correction: A Survey of the State of the Art,

C. Bryant, Z. Yuan, M. R. Qorib, H. Cao, H. T. Ng, and T. Briscoe, “Grammatical Error Correction: A Survey of the State of the Art,” Computational Linguistics, vol. 49, no. 3, pp. 643–701, 09 2023

work page 2023

[7] [7]

The CoNLL-2014 Shared Task on Grammatical Error Correction,

H. T. Ng, S. M. Wu, T. Briscoe, C. Hadiwinoto, R. H. Susanto, and C. Bryant, “The CoNLL-2014 Shared Task on Grammatical Error Correction,” in Proceedings of the 18th conference on computational natural language learning: shared task, 2014, pp. 1–14. [Online]. Available: https://aclanthology.org/W14-1701/

work page 2014

[8] [8]

The BEA- 2019 Shared Task on Grammatical Error Correction,

C. Bryant, M. Felice, Ø. E. Andersen, and T. Briscoe, “The BEA- 2019 Shared Task on Grammatical Error Correction,” inProceed- ings of the fourteenth workshop on innovative use of NLP for building educational applications, 2019, pp. 52–75

work page 2019