The Impact of Editorial Intervention on Detecting Native Language Traces

Ahmet Yavuz Uluslu; Gerold Schneider; Kate Knill; Mark Gales

arxiv: 2605.10216 · v1 · submitted 2026-05-11 · 💻 cs.CL

The Impact of Editorial Intervention on Detecting Native Language Traces

Ahmet Yavuz Uluslu , Mark Gales , Kate Knill , Gerold Schneider This is my paper

Pith reviewed 2026-05-12 03:58 UTC · model grok-4.3

classification 💻 cs.CL

keywords native language identificationgrammatical error correctionparaphrasingL1 attributioneditorial interventionlinguistic tracesAI writing assistance

0 comments

The pith

Native language traces survive light AI edits but vanish under fluency paraphrasing.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tests how much an author's native language remains detectable in English essays after different strengths of AI editing. It runs 450 essays through minimal corrections, fluency improvements, and full paraphrasing, then measures how well standard NLI models still identify the original L1. The results show that models do not depend only on obvious grammar mistakes. Deeper patterns such as odd word choices, pragmatic habits, and cultural viewpoints stay visible after light fixes. Once the text is rewritten to sound fully natural, those patterns are smoothed away and accuracy falls sharply.

Core claim

L1 attribution does not entirely depend on surface-level errors. The detection models instead leverage deeper L1 features such as unidiomatic lexico-semantic choices, pragmatic transfer, and the author's underlying cultural perspective. Minimal edits preserve these structural traces and maintain high profiling accuracy. In contrast, fluency edits and paraphrasing normalize these L1 features, leading to a severe degradation in performance.

What carries the argument

The graded editorial-intervention pipeline that applies increasing levels of grammatical error correction and paraphrasing to the same essays before re-testing NLI accuracy.

If this is right

Light AI corrections will leave native-language signals largely intact for current detectors.
Heavy fluency rewriting will make reliable L1 attribution much harder.
NLI systems already exploit lexico-semantic and pragmatic patterns rather than error lists alone.
Texts produced in human-AI collaboration can still carry detectable background information unless the AI rewrites aggressively.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

AI writing tools may unintentionally preserve or erase cultural identity markers based on how deeply they edit.
Privacy risks exist even in lightly corrected non-native writing because deeper traces remain.
Detection systems could be made more robust by training explicitly on these deeper features instead of surface errors.

Load-bearing premise

The specific strengths of GEC and paraphrasing used here match the kinds of edits real AI tools make and the tested NLI models are not overfitted to this one essay collection.

What would settle it

Running the same NLI models on the Write & Improve essays after full paraphrasing and finding accuracy stays near the original level would show that deeper L1 features are not actually being removed.

Figures

Figures reproduced from arXiv: 2605.10216 by Ahmet Yavuz Uluslu, Gerold Schneider, Kate Knill, Mark Gales.

read the original abstract

Native Language Identification (NLI) is the task of determining an author's native language (L1) from their non-native writings. With the advent of human-AI co-authorship, non-native texts are routinely corrected and rewritten by large language models, fundamentally altering the linguistic features NLI models depend on. In this paper, we investigate the robustness of L1 traces across increasing degrees of editorial intervention. By processing 450 essays from the Write & Improve 2024 corpus through varying levels of grammatical error correction (GEC) and paraphrasing, we demonstrate that L1 attribution does not entirely depend on surface-level errors. Instead, the detection models leverage deeper L1 features: unidiomatic lexico-semantic choices, pragmatic transfer, and the author's underlying cultural perspective. We find that minimal edits preserve these structural traces and maintain high profiling accuracy. In contrast, fluency edits and paraphrasing normalize these L1 features, leading to a severe degradation in performance.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Light grammatical fixes leave NLI signals mostly intact while fluency edits and paraphrasing wipe them out, but the paper still needs the actual numbers and checks to make the claim reliable.

read the letter

The main thing to know is that the authors ran the same 450 Write & Improve essays through graded levels of editing and found that minimal GEC keeps native-language detection accuracy high, but once the interventions reach fluency correction or full paraphrasing the models lose the signal. That directional split between surface fixes and deeper normalization is the concrete result worth noting. The setup itself is straightforward and useful: they hold the source texts fixed and vary only the intervention strength, which lets them separate surface error removal from changes to lexico-semantic and pragmatic choices. That framing is cleaner than most prior NLI robustness work I have seen. They also correctly flag that real-world LLM assistance now routinely goes beyond simple grammar fixes, so the question matters for anyone building or using authorship tools. The soft spots are exactly where the stress-test note points. No accuracy figures, no model specifications, no statistical tests, and no error analysis appear in the abstract, so it is impossible to judge how large the drop actually is or whether the models were already tuned to this narrow corpus. The chosen GEC and paraphrasing prompts also need explicit validation against what users actually do with current LLMs; if the heavy-edit condition is stronger or weaker than typical practice, the claimed robustness boundary shifts. The deeper-feature claim is plausible on its face but currently rests on that unverified operationalization. This is the sort of paper that belongs in an applied NLP or digital-forensics venue. Readers who work on NLI or forensic text analysis will want to see the full tables and the prompt details even if they end up disagreeing with the strength of the conclusion. It is solid enough to send to referees; the experiment design is simple and the question is timely, so a review can focus on tightening the methods and adding the missing quantitative evidence rather than rejecting the premise outright.

Referee Report

2 major / 2 minor

Summary. The paper investigates the robustness of Native Language Identification (NLI) models to AI-driven editorial interventions. Using 450 essays from the Write & Improve 2024 corpus, the authors apply varying levels of grammatical error correction (GEC) and paraphrasing, then evaluate how these transformations affect L1 attribution accuracy. The central claim is that L1 traces are not limited to surface-level errors; minimal edits preserve deeper features (unidiomatic lexico-semantic choices, pragmatic transfer, and cultural perspective), maintaining high profiling accuracy, while fluency edits and paraphrasing normalize these features and cause severe performance degradation.

Significance. If the empirical findings hold after addressing methodological details, the work makes a timely contribution to computational linguistics by showing that NLI signals are partially resilient to light editing but vulnerable to heavier AI intervention. This has practical implications for authorship attribution, forensic linguistics, and detection of AI co-authorship. The use of a real non-native corpus and controlled intervention levels is a strength, providing falsifiable evidence on feature robustness rather than relying solely on theoretical arguments.

major comments (2)

[Methods] Methods section: The manuscript provides insufficient detail on the NLI models (architectures, pre-training corpora, fine-tuning procedures, or hyper-parameters). Without this, it is impossible to assess whether the observed retention of accuracy after minimal edits reflects genuine deeper L1 features or corpus-specific overfitting to the 450-essay Write & Improve subset, directly undermining the central claim that models leverage structural traces beyond surface errors.
[Experimental Setup] Experimental design (GEC and paraphrasing pipeline): The paper does not specify the exact prompts, underlying LLMs, or quantitative thresholds (e.g., edit distance, fluency scores) used to operationalize 'minimal' vs. 'fluency' edits. This makes it difficult to verify whether the chosen intervention levels are representative of real-world AI editorial practices, as required to support the claim that paraphrasing normalizes L1 features while minimal edits do not.

minor comments (2)

[Abstract] Abstract: Include at least one key quantitative result (e.g., accuracy drop percentages or statistical significance) to convey the magnitude of the observed degradation.
[Introduction] Notation: Define 'NLI models' more explicitly on first use and clarify whether they are zero-shot or fine-tuned on the corpus.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed feedback on our manuscript. We agree that greater methodological transparency is required to support the central claims and enable reproducibility. We will revise the paper to address both major comments by expanding the relevant sections with the requested details. Point-by-point responses follow.

read point-by-point responses

Referee: [Methods] Methods section: The manuscript provides insufficient detail on the NLI models (architectures, pre-training corpora, fine-tuning procedures, or hyper-parameters). Without this, it is impossible to assess whether the observed retention of accuracy after minimal edits reflects genuine deeper L1 features or corpus-specific overfitting to the 450-essay Write & Improve subset, directly undermining the central claim that models leverage structural traces beyond surface errors.

Authors: We agree that the current Methods section lacks sufficient detail on the NLI models, which limits assessment of potential overfitting versus genuine feature retention. In the revised manuscript, we will add a dedicated subsection specifying the model architectures (transformer-based models such as XLM-RoBERTa), pre-training corpora, fine-tuning procedures on the Write & Improve 2024 data, hyper-parameters, and cross-validation strategy. We will also include an analysis comparing performance on held-out data and baseline models to support that the retained accuracy after minimal edits reflects deeper L1 traces rather than subset-specific overfitting. This revision directly addresses the concern and strengthens the central claim. revision: yes
Referee: [Experimental Setup] Experimental design (GEC and paraphrasing pipeline): The paper does not specify the exact prompts, underlying LLMs, or quantitative thresholds (e.g., edit distance, fluency scores) used to operationalize 'minimal' vs. 'fluency' edits. This makes it difficult to verify whether the chosen intervention levels are representative of real-world AI editorial practices, as required to support the claim that paraphrasing normalizes L1 features while minimal edits do not.

Authors: We acknowledge that the experimental setup description is insufficiently specific regarding the GEC and paraphrasing pipeline. In the revised manuscript, we will provide the exact prompts used, identify the underlying LLMs (including version and access details), and report quantitative thresholds such as edit-distance metrics and fluency scores that define the 'minimal', 'fluency', and 'paraphrasing' intervention levels. We will also add representative examples of each edit type to demonstrate alignment with real-world AI editorial practices. These additions will allow verification of the intervention levels and better support the claim regarding differential impact on L1 feature preservation. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical comparison of NLI performance on transformed texts

full rationale

The paper reports an empirical study applying GEC and paraphrasing transformations to 450 Write & Improve essays and measuring resulting NLI accuracy. No equations, fitted parameters, self-definitional claims, or load-bearing self-citations appear in the abstract or described methodology. Claims about preservation of deeper L1 features rest on direct experimental contrasts rather than any derivation that reduces to its own inputs by construction. This is a standard self-contained empirical evaluation against external benchmarks (fixed NLI models and corpus).

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that the chosen GEC and paraphrasing operations are representative of real LLM editing and that standard NLI models capture the relevant L1 signals.

axioms (1)

domain assumption Standard NLI models trained on unedited text can be applied directly to edited versions without retraining or domain adaptation.
The experiment applies existing models to transformed text without mentioning retraining.

pith-pipeline@v0.9.0 · 5468 in / 1117 out tokens · 60835 ms · 2026-05-12T03:58:37.622256+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages

[1]

The Bell System Technical Journal , volume=

A mathematical theory of communication , author=. The Bell System Technical Journal , volume=. 1948 , url =

work page 1948
[2]

Diane Nicholls and Andrew Caines and Paula Buttery , year =. The

work page
[3]

Weighting Test Samples in

Blanchard, Daniel and Tetreault, Joel and Higgins, Derrick and Cahill, Aoife and Chodorow, Martin , journal=. 2013 , doi =. https://onlinelibrary.wiley.com/doi/pdf/10.1002/j.2333-8504.2013.tb02331.x , publisher=

work page doi:10.1002/j.2333-8504.2013.tb02331.x 2013
[4]

Language Resources and Evaluation , volume=

Authorship attribution in the wild , author=. Language Resources and Evaluation , volume=. 2011 , publisher=

work page 2011
[5]

Responsible guidelines for authorship attribution tasks in

Saxena, Vageesh and Tam. Responsible guidelines for authorship attribution tasks in. Ethics and Information Technology , volume=. 2025 , publisher=

work page 2025
[6]

doi:10.18653/v1/2024.eacl-long.8 , pages=

Liusie, Adian and Manakul, Potsawee and Gales, Mark , booktitle=. doi:10.18653/v1/2024.eacl-long.8 , pages=

work page doi:10.18653/v1/2024.eacl-long.8 2024
[7]

Advances in Neural Information Processing Systems , pages=

Can you trust your model's uncertainty? Evaluating predictive uncertainty under dataset shift , author=. Advances in Neural Information Processing Systems , pages=

work page
[8]

arXiv preprint arXiv:2601.18056 , year=

Neurocomputational mechanisms of syntactic transfer in bilingual sentence production , author=. arXiv preprint arXiv:2601.18056 , year=

work page arXiv
[9]

Natural Language Engineering , volume=

Exploiting native language interference for native language identification , author=. Natural Language Engineering , volume=. 2022 , publisher=

work page 2022
[10]

Digital Scholarship in the Humanities , volume=

Unravelling interlanguage facts via explainable machine learning , author=. Digital Scholarship in the Humanities , volume=. 2023 , publisher=

work page 2023
[11]

Findings of the Association for Computational Linguistics: ACL 2025 , pages=

Machine translation models are zero-shot detectors of translation direction , author=. Findings of the Association for Computational Linguistics: ACL 2025 , pages=

work page 2025
[12]

Policing: A Journal of Policy and Practice , volume=

The application of forensic linguistics in cybercrime investigations , author=. Policing: A Journal of Policy and Practice , volume=. 2021 , publisher=

work page 2021
[13]

Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , pages=

Robust native language identification through agentic decomposition , author=. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , pages=

work page 2025
[14]

A report on the

Malmasi, Shervin and Evanini, Keelan and Cahill, Aoife and Tetreault, Joel and Pugh, Robert and Hamill, Christopher and Napolitano, Diane and Qian, Yao , booktitle=. A report on the. doi:10.18653/v1/W17-5007

work page doi:10.18653/v1/w17-5007
[15]

Authorship attribution in the era of

Huang, Baixiang and Chen, Canyu and Shu, Kai , journal=. Authorship attribution in the era of. 2025 , publisher=

work page 2025
[16]

https://aclanthology.org/2025.naacl-srw.19/

Multilingual native language identification with large language models , author=. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 4: Student Research Workshop) , url = "https://aclanthology.org/2025.naacl-srw.19/", doi = "10.18653/v1/2025.naacl-...

work page doi:10.18653/v1/2025.naacl-srw.19 2025
[17]

arXiv preprint arXiv:2312.07819 , year=

Native language identification with large language models , author=. arXiv preprint arXiv:2312.07819 , year=

work page arXiv
[18]

Next-generation phishing: How

Afane, Khalifa and Wei, Wenqi and Mao, Ying and Farooq, Junaid and Chen, Juntao , booktitle=. Next-generation phishing: How. 2024 , organization=

work page 2024
[19]

Chen, Fengchao and Wu, Tingmin and Nguyen, Van and Rudolph, Carsten , journal=

work page
[20]

doi:10.18653/v1/2024.naacl-long.173

Goswami, Dhiman and Thilagan, Sharanya and North, Kai and Malmasi, Shervin and Zampieri, Marcos , booktitle=. doi:10.18653/v1/2024.naacl-long.173

work page doi:10.18653/v1/2024.naacl-long.173 2024
[21]

Web-browsing

Alizadeh, Meysam and Gilardi, Fabrizio and Samei, Zeynab and Mosleh, Mohsen , journal=. Web-browsing

work page
[22]

Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024) , pages=

Pseudonymization categories across domain boundaries , author=. Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024) , pages=

work page 2024
[23]

Proceedings of the 12th Workshop on NLP for Similar Languages, Varieties and Dialects , pages=

Leveraging open-source large language models for native language identification , author=. Proceedings of the 12th Workshop on NLP for Similar Languages, Varieties and Dialects , pages=

work page
[24]

Investigating linguistic abilities of

Uluslu, Ahmet Yavuz and Schneider, Gerold. Investigating linguistic abilities of. Proceedings of the 14th Workshop on Natural Language Processing for Computer Assisted Language Learning. 2025

work page 2025
[25]

Elizabeth and Goswami, Dhiman and North, Kai and Zampieri, Marcos and Anastasopoulos, Antonios

Acharya, Poorvi and Liebl, J. Elizabeth and Goswami, Dhiman and North, Kai and Zampieri, Marcos and Anastasopoulos, Antonios. Tracing. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025. doi:10.18653/v1/2025.emnlp-main.766

work page doi:10.18653/v1/2025.emnlp-main.766 2025
[26]

Prompting open-source and commercial language models for grammatical error correction of

Davis, Christopher and Caines, Andrew and Andersen,. Prompting open-source and commercial language models for grammatical error correction of. Findings of the association for computational linguistics: ACL 2024 , pages=

work page 2024
[27]

Dentella, Vittoria and Huang, Weihang and Mansi, Silvia Angela and Grieve, Jack and Leivada, Evelina , journal=

work page
[28]

Automatic authorship analysis in human-

Richburg, Aquia and Bao, Calvin and Carpuat, Marine , booktitle=. Automatic authorship analysis in human-

work page
[29]

2024 , publisher=

Yang, Lu and Li, Rui , journal=. 2024 , publisher=

work page 2024
[30]

Computational Linguistics , pages=

Grammatical error correction: A survey of the state of the art , author=. Computational Linguistics , pages=

work page
[31]

Zahid, Iqra and Sun, Youcheng and Batista-Navarro, Riza Theresa , booktitle=

work page
[32]

Adapting

Staruch, Ryszard and Gralinski, Filip and Dzienisiewicz, Daniel. Adapting. Proceedings of the 20th Workshop on Innovative Use of NLP for Building Educational Applications (BEA 2025). 2025. doi:10.18653/v1/2025.bea-1.9

work page doi:10.18653/v1/2025.bea-1.9 2025
[33]

Introducing

Kovalchuk, Roman and Romanyshyn, Mariana and Ivaniuk, Petro. Introducing. Proceedings of the Fourth Ukrainian Natural Language Processing Workshop (UNLP 2025). 2025. doi:10.18653/v1/2025.unlp-1.17

work page doi:10.18653/v1/2025.unlp-1.17 2025
[34]

Zhan, Yuhao and Zhang, Yuqing and Yuan, Jing and Ma, Qixiang and Yang, Zhiqi and Gu, Yu and Liu, Zemin and Wu, Fei , booktitle=

work page
[35]

Proceedings of the 2012 conference of the north american chapter of the association for computational linguistics: human language technologies , pages=

Better evaluation for grammatical error correction , author=. Proceedings of the 2012 conference of the north american chapter of the association for computational linguistics: human language technologies , pages=

work page 2012
[36]

Zhang, Tianyi and Kishore, Varsha and Wu, Felix and Weinberger, Kilian Q and Artzi, Yoav , journal=

work page
[37]

Introducing the

Knill, Kate M and Nicholls, Diane and Gales, Mark JF and Qian, Mengjie and Stroinski, Pawel , booktitle=. Introducing the. doi:10.21437/SLaTE.2025-34 , year=

work page doi:10.21437/slate.2025-34 2025
[38]

Catch Me If You Can? Not Yet:

Wang, Zhengxiang and Tripto, Nafis Irtiza and Park, Solha and Li, Zhenzhen and Zhou, Jiawei , booktitle=. Catch Me If You Can? Not Yet:

work page
[39]

Proceedings of the 55th annual meeting of the association for computational linguistics (Volume 1: Long Papers) , pages=

Automatic annotation and evaluation of error types for grammatical error correction , author=. Proceedings of the 55th annual meeting of the association for computational linguistics (Volume 1: Long Papers) , pages=

work page
[40]

Bilingualism: Language and cognition , volume=

Conceptual transfer: Crosslinguistic effects in categorization and construal , author=. Bilingualism: Language and cognition , volume=. 2011 , publisher=

work page 2011
[41]

Linguistics and philosophy , volume=

Number marking and (in) definiteness in kind terms , author=. Linguistics and philosophy , volume=. 2004 , publisher=

work page 2004

[1] [1]

The Bell System Technical Journal , volume=

A mathematical theory of communication , author=. The Bell System Technical Journal , volume=. 1948 , url =

work page 1948

[2] [2]

Diane Nicholls and Andrew Caines and Paula Buttery , year =. The

work page

[3] [3]

Weighting Test Samples in

Blanchard, Daniel and Tetreault, Joel and Higgins, Derrick and Cahill, Aoife and Chodorow, Martin , journal=. 2013 , doi =. https://onlinelibrary.wiley.com/doi/pdf/10.1002/j.2333-8504.2013.tb02331.x , publisher=

work page doi:10.1002/j.2333-8504.2013.tb02331.x 2013

[4] [4]

Language Resources and Evaluation , volume=

Authorship attribution in the wild , author=. Language Resources and Evaluation , volume=. 2011 , publisher=

work page 2011

[5] [5]

Responsible guidelines for authorship attribution tasks in

Saxena, Vageesh and Tam. Responsible guidelines for authorship attribution tasks in. Ethics and Information Technology , volume=. 2025 , publisher=

work page 2025

[6] [6]

doi:10.18653/v1/2024.eacl-long.8 , pages=

Liusie, Adian and Manakul, Potsawee and Gales, Mark , booktitle=. doi:10.18653/v1/2024.eacl-long.8 , pages=

work page doi:10.18653/v1/2024.eacl-long.8 2024

[7] [7]

Advances in Neural Information Processing Systems , pages=

Can you trust your model's uncertainty? Evaluating predictive uncertainty under dataset shift , author=. Advances in Neural Information Processing Systems , pages=

work page

[8] [8]

arXiv preprint arXiv:2601.18056 , year=

Neurocomputational mechanisms of syntactic transfer in bilingual sentence production , author=. arXiv preprint arXiv:2601.18056 , year=

work page arXiv

[9] [9]

Natural Language Engineering , volume=

Exploiting native language interference for native language identification , author=. Natural Language Engineering , volume=. 2022 , publisher=

work page 2022

[10] [10]

Digital Scholarship in the Humanities , volume=

Unravelling interlanguage facts via explainable machine learning , author=. Digital Scholarship in the Humanities , volume=. 2023 , publisher=

work page 2023

[11] [11]

Findings of the Association for Computational Linguistics: ACL 2025 , pages=

Machine translation models are zero-shot detectors of translation direction , author=. Findings of the Association for Computational Linguistics: ACL 2025 , pages=

work page 2025

[12] [12]

Policing: A Journal of Policy and Practice , volume=

The application of forensic linguistics in cybercrime investigations , author=. Policing: A Journal of Policy and Practice , volume=. 2021 , publisher=

work page 2021

[13] [13]

Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , pages=

Robust native language identification through agentic decomposition , author=. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , pages=

work page 2025

[14] [14]

A report on the

Malmasi, Shervin and Evanini, Keelan and Cahill, Aoife and Tetreault, Joel and Pugh, Robert and Hamill, Christopher and Napolitano, Diane and Qian, Yao , booktitle=. A report on the. doi:10.18653/v1/W17-5007

work page doi:10.18653/v1/w17-5007

[15] [15]

Authorship attribution in the era of

Huang, Baixiang and Chen, Canyu and Shu, Kai , journal=. Authorship attribution in the era of. 2025 , publisher=

work page 2025

[16] [16]

https://aclanthology.org/2025.naacl-srw.19/

Multilingual native language identification with large language models , author=. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 4: Student Research Workshop) , url = "https://aclanthology.org/2025.naacl-srw.19/", doi = "10.18653/v1/2025.naacl-...

work page doi:10.18653/v1/2025.naacl-srw.19 2025

[17] [17]

arXiv preprint arXiv:2312.07819 , year=

Native language identification with large language models , author=. arXiv preprint arXiv:2312.07819 , year=

work page arXiv

[18] [18]

Next-generation phishing: How

Afane, Khalifa and Wei, Wenqi and Mao, Ying and Farooq, Junaid and Chen, Juntao , booktitle=. Next-generation phishing: How. 2024 , organization=

work page 2024

[19] [19]

Chen, Fengchao and Wu, Tingmin and Nguyen, Van and Rudolph, Carsten , journal=

work page

[20] [20]

doi:10.18653/v1/2024.naacl-long.173

Goswami, Dhiman and Thilagan, Sharanya and North, Kai and Malmasi, Shervin and Zampieri, Marcos , booktitle=. doi:10.18653/v1/2024.naacl-long.173

work page doi:10.18653/v1/2024.naacl-long.173 2024

[21] [21]

Web-browsing

Alizadeh, Meysam and Gilardi, Fabrizio and Samei, Zeynab and Mosleh, Mohsen , journal=. Web-browsing

work page

[22] [22]

Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024) , pages=

Pseudonymization categories across domain boundaries , author=. Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024) , pages=

work page 2024

[23] [23]

Proceedings of the 12th Workshop on NLP for Similar Languages, Varieties and Dialects , pages=

Leveraging open-source large language models for native language identification , author=. Proceedings of the 12th Workshop on NLP for Similar Languages, Varieties and Dialects , pages=

work page

[24] [24]

Investigating linguistic abilities of

Uluslu, Ahmet Yavuz and Schneider, Gerold. Investigating linguistic abilities of. Proceedings of the 14th Workshop on Natural Language Processing for Computer Assisted Language Learning. 2025

work page 2025

[25] [25]

Elizabeth and Goswami, Dhiman and North, Kai and Zampieri, Marcos and Anastasopoulos, Antonios

Acharya, Poorvi and Liebl, J. Elizabeth and Goswami, Dhiman and North, Kai and Zampieri, Marcos and Anastasopoulos, Antonios. Tracing. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025. doi:10.18653/v1/2025.emnlp-main.766

work page doi:10.18653/v1/2025.emnlp-main.766 2025

[26] [26]

Prompting open-source and commercial language models for grammatical error correction of

Davis, Christopher and Caines, Andrew and Andersen,. Prompting open-source and commercial language models for grammatical error correction of. Findings of the association for computational linguistics: ACL 2024 , pages=

work page 2024

[27] [27]

Dentella, Vittoria and Huang, Weihang and Mansi, Silvia Angela and Grieve, Jack and Leivada, Evelina , journal=

work page

[28] [28]

Automatic authorship analysis in human-

Richburg, Aquia and Bao, Calvin and Carpuat, Marine , booktitle=. Automatic authorship analysis in human-

work page

[29] [29]

2024 , publisher=

Yang, Lu and Li, Rui , journal=. 2024 , publisher=

work page 2024

[30] [30]

Computational Linguistics , pages=

Grammatical error correction: A survey of the state of the art , author=. Computational Linguistics , pages=

work page

[31] [31]

Zahid, Iqra and Sun, Youcheng and Batista-Navarro, Riza Theresa , booktitle=

work page

[32] [32]

Adapting

Staruch, Ryszard and Gralinski, Filip and Dzienisiewicz, Daniel. Adapting. Proceedings of the 20th Workshop on Innovative Use of NLP for Building Educational Applications (BEA 2025). 2025. doi:10.18653/v1/2025.bea-1.9

work page doi:10.18653/v1/2025.bea-1.9 2025

[33] [33]

Introducing

Kovalchuk, Roman and Romanyshyn, Mariana and Ivaniuk, Petro. Introducing. Proceedings of the Fourth Ukrainian Natural Language Processing Workshop (UNLP 2025). 2025. doi:10.18653/v1/2025.unlp-1.17

work page doi:10.18653/v1/2025.unlp-1.17 2025

[34] [34]

Zhan, Yuhao and Zhang, Yuqing and Yuan, Jing and Ma, Qixiang and Yang, Zhiqi and Gu, Yu and Liu, Zemin and Wu, Fei , booktitle=

work page

[35] [35]

Proceedings of the 2012 conference of the north american chapter of the association for computational linguistics: human language technologies , pages=

Better evaluation for grammatical error correction , author=. Proceedings of the 2012 conference of the north american chapter of the association for computational linguistics: human language technologies , pages=

work page 2012

[36] [36]

Zhang, Tianyi and Kishore, Varsha and Wu, Felix and Weinberger, Kilian Q and Artzi, Yoav , journal=

work page

[37] [37]

Introducing the

Knill, Kate M and Nicholls, Diane and Gales, Mark JF and Qian, Mengjie and Stroinski, Pawel , booktitle=. Introducing the. doi:10.21437/SLaTE.2025-34 , year=

work page doi:10.21437/slate.2025-34 2025

[38] [38]

Catch Me If You Can? Not Yet:

Wang, Zhengxiang and Tripto, Nafis Irtiza and Park, Solha and Li, Zhenzhen and Zhou, Jiawei , booktitle=. Catch Me If You Can? Not Yet:

work page

[39] [39]

Proceedings of the 55th annual meeting of the association for computational linguistics (Volume 1: Long Papers) , pages=

Automatic annotation and evaluation of error types for grammatical error correction , author=. Proceedings of the 55th annual meeting of the association for computational linguistics (Volume 1: Long Papers) , pages=

work page

[40] [40]

Bilingualism: Language and cognition , volume=

Conceptual transfer: Crosslinguistic effects in categorization and construal , author=. Bilingualism: Language and cognition , volume=. 2011 , publisher=

work page 2011

[41] [41]

Linguistics and philosophy , volume=

Number marking and (in) definiteness in kind terms , author=. Linguistics and philosophy , volume=. 2004 , publisher=

work page 2004