Pith. sign in

REVIEW 4 major objections 4 minor 36 references

Digital Guardians: Can GPT-4, Perspective API, and Moderation API reliably detect hate speech in reader comments of German online newspapers?

T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper reports that GPT-4o, prompted with a single example, detects hate speech in German reader comments better than a fine-tuned BERT baseline and two commercial moderation APIs.

desk verdict A useful point-in-time benchmark of GPT-4o vs. two moderation APIs on German news comments, but the five-point lead lacks significance testing and the reannotation results are biased in GPT-4o's favor. read the letter →

arxiv 2501.01256 v1 pith:7N6EKHEZ submitted 2025-01-02 cs.CL cs.LG

Manuel Weber , Moritz Huber , Maximilian Auch , Alexander Döschl , Max-Emanuel Keller , Peter Mandl This is my paper · ORCID
classification cs.CLcs.LG
keywords hatespeechdetectionlargelanguagemodelsGPT-4oPerspectiveAPIModerationHOCON34kGermanNLPannotationquality
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper asks whether commercial AI moderation services can replace or supplement human moderators for hate speech in German online newspaper comments. On a 1,592-comment test set from the HOCON34k corpus, the authors compare prompted GPT-4o (zero-, one-, and few-shot), Google's Perspective API, and OpenAI's Moderation API against a fine-tuned BERT baseline. They report that GPT-4o with one-shot prompting reaches a Champion-Challenger score S = 0.7059, about five percentage points above the baseline's S = 0.652, while both moderation APIs fall short. The paper also reannotates a large share of the test set, finds roughly 64% of the reviewed labels were changed, and shows that correcting them improves nearly every system, most of all GPT-4o. The authors conclude that a general-purpose LLM can outperform task-specific models on this task, while cautioning that their reannotation procedure may favor GPT.

What carries the argument

The argument runs through three components: the HOCON34k corpus and its 1,592-sample test set; the combined score S = ((MCC+1)/2 + F2)/2, which rewards both balanced prediction and recall; and the prompted evaluation of GPT-4o, whose prompt encodes the corpus's hate-speech definition and optionally one or four labeled examples. The reannotation study additionally uses GPT-4o's zero-shot majority votes to select 314 labels for manual re-review.

What would settle it

Run a fully independent reannotation of all 1,592 test comments by annotators who never see GPT-4o's predictions, then recompute every system's S score on the corrected labels; if the BERT baseline's S equals or exceeds GPT-4o's, the claimed five-point advantage is an artifact.

Watch

Extended reading notes

Core claim

The paper's central claim is that GPT-4o, a general-purpose large language model used without any fine-tuning, detects hate speech in German reader comments more reliably than both a commercial moderation API from Google and one from OpenAI, and also exceeds the published HOCON34k BERT baseline. The best variant, GPT-4o with one example in the prompt, achieved an S score of 0.7059, where S averages the normalized Matthews correlation coefficient and the F2 score, compared with the baseline's 0.652 on the original 1,592-sample test set. After reannotating 314 labels flagged by comparing GPT-4o's output with the original annotations, the same variant reaches S = 0.8268 versus the baseline's 0.723, widening the gap. The authors also show that the Perspective API, at its recommended high threshold, misses almost all hate comments (only 6 of 329 detected), while the Moderation API sits in between.

Load-bearing premise

The original HOCON34k test labels are correct enough to serve as ground truth for comparing the systems.

Editorial extensions

If this is right

  • GPT-4o with one-shot prompting reaches S = 0.7059 on the original 1,592-sample test set, roughly five percentage points above the HOCON34k baseline's S = 0.652, while the Moderation API (S = 0.6327) and the best Perspective API configuration (S = 0.5613) trail behind.
  • The Perspective API's low false-positive count at tau = 0.8 (only 1 false positive in 1,592 samples) comes at the cost of missing 323 of 329 hate-speech comments, making it unsuitable for recall-oriented moderation.
  • After reannotating 314 labels, 201 labels changed (64 from hate to non-hate and 137 from non-hate to hate), raising the hate-speech share from 20.7% to 25.2%; every system improves on the corrected set, with GPT-4o One-Shot reaching S = 0.8268.
  • Few-shot prompting with four examples did not outperform one-shot or zero-shot on the original test set, suggesting the in-context examples add little for this task.

Reading between the lines

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

  • Because the reannotation only reviewed samples GPT-4o flagged as misclassified, the post-reannotation S gain for GPT-4o is partly mechanical; an independent full reannotation would likely show a smaller and fairer lead.
  • The S metric averages a normalized MCC and F2, weighting recall twice as much as precision; under a precision-biased metric like F0.5, the ranking between GPT-4o and the baseline could change.
  • The results suggest a general-purpose LLM API could serve as a drop-in moderation pre-screen for German newsrooms, but real deployments would need to test context-dependent comments, since the dataset deliberately omitted conversational context.
  • A natural next experiment is to measure the same systems on the reannotated set with the full comment thread and article context included, which the authors flag as future work.
Share X Bluesky LinkedIn Reddit HN

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 compares GPT-4o (Zero-Shot, One-Shot, Few-Shot), Jigsaw's Perspective API at three thresholds, and OpenAI's Moderation API on the HOCON34k test set of 1,592 German newspaper comments, using the combined Champion-Challenger score S=(M_norm+F2)/2 as the primary metric. On the original labels, GPT-4o One-Shot achieves S=0.7059 versus the imported HOCON34k BERT baseline S=0.652, a difference of about 0.054; the paper concludes that GPT-4o outperforms both commercial APIs and the fine-tuned baseline. A GPT-selected reannotation of part of the test set is then used to claim that GPT-4o's advantage over the baseline nearly doubles on corrected labels.

Significance. If the result holds, it is practically relevant: a prompted general-purpose LLM beating a fine-tuned German BERT classifier and two commercial moderation APIs on a real-world German newspaper comment benchmark would be an actionable finding for online moderation workflows. The manuscript has genuine strengths: it reports three repeated runs for each system, gives confusion matrices for representative runs, specifies the exact GPT-4o snapshot (gpt-4o-2024-05-13), shows the full prompt, and is candid about non-determinism and label-error concerns. The main weakness is that the headline comparison rests on potentially noisy labels and an imported baseline, and the reannotation experiment is biased in favor of GPT-4o because the review set is selected by GPT-4o's disagreements. These issues are fixable, but the current evidence does not fully establish the abstract's unqualified superiority claim.

major comments (4)
  1. [§4.1, Table 2; §5.1] The headline 0.0539 S advantage over the HOCON34k baseline is not robust to the label noise the paper itself documents. Section 5.1 projects that 78–96 of the 1,592 test labels (4.9–6.0%) may be incorrect, yet every S, F2, and MCC value in Table 2 is computed against those labels, and the baseline threshold 0.523 is defined on the same labels. A few dozen label corrections can shift S by more than the observed gap, so the ranking could change under corrected ground truth. The paper concedes this in §4.2 ('Eliminating potentially erroneous data might lead to different results'). An unbiased reannotation of a random sample, or a sensitivity analysis that re-evaluates all systems under plausible label-correction scenarios, is needed before the superiority claim can be accepted.
  2. [§5.2, §5.3] The reannotation experiment is circular with respect to GPT-4o's evaluation. Section 5.2 uses GPT-4o Zero-Shot majority decisions to select 314 samples for manual review; 201 labels are changed, 64% in line with GPT-4o's suggestion. Section 5.3 then reports improved S scores for GPT-4o on the revised set (One-Shot rising from 0.7059 to 0.8268) and states that the advantage over the baseline nearly doubled. Because the review set was chosen by GPT-4o's disagreements, the revised labels are not an unbiased estimate of true labels, and the results in Table 5 and Figure 2 cannot be used as evidence of GPT-4o's superiority. The manuscript itself acknowledges this in the final paragraph of §5.3 ('the selective review may introduce a bias in favor of GPT'). I recommend either removing the §5.3 comparison from the central claims or re-running the evaluation on a random sample reannotated without GPT-based selection.
  3. [§3.1, Table 2] The baseline comparison is not made under a common protocol. Section 3.1 states that the HOCON34k baseline results 'were taken from Keller et al. (2024) for comparison,' and the threshold 0.523 is not recomputed in this study. Since the APIs were queried in June 2024 and the baseline may have been produced with different preprocessing or threshold selection, the 0.0539 S difference (0.7059 vs 0.652) could reflect setup variation rather than model quality. The authors demonstrate access to the baseline classifier in §5.3, where they rerun it with threshold 0.557 on reannotated data, so the baseline should also be rerun on the original test split under identical conditions, and this recomputed value should replace the imported one.
  4. [§4.1, Table 2] No significance tests or confidence intervals are reported for any of the differences. With only three repeated runs, the differences among GPT-4o variants (S = 0.6908, 0.7059, 0.6897) and the 0.0539 gap over the baseline are not shown to be statistically reliable; the same applies to the comparisons against the APIs. I would like to see bootstrap confidence intervals on S, F2, and MCC, or a paired test across the 1,592 samples, before the abstract's 'outperforms' claim is asserted.
minor comments (4)
  1. [§5.3] Figure 2 is referenced in the text but the figure content is not visible in the manuscript; please ensure the figure is actually included.
  2. [Table 2] The baseline column is labeled 'τ = 0.523', but the HOCON34k baseline is a fine-tuned BERT classifier with a probability threshold, not an API parameter; the notation should be clarified in the caption.
  3. [§5.1] The projection of 78–96 erroneous labels is derived from a reannotation of only 91 texts; please report the exact confidence-interval calculation and the per-annotator counts, because the current statement ('error range 10%, confidence interval 90%') is not self-contained.
  4. [§3.6] The paper lists three Perspective API thresholds without stating whether they were fixed before seeing the test results or chosen after inspection; a sentence on the threshold-selection rule would make the comparison cleaner.

Circularity Check

1 steps flagged · score 4.0 of 10

Self-acknowledged GPT-selected reannotation loop inflates the secondary 'improved results,' while the headline comparison on the original HOCON34k labels remains an independent external evaluation.

  1. fitted input called prediction [Section 5.2 (Reannotation of Testdata) and Section 5.3 (Improved Results), Tables 4–6 and Fig. 2]
    "However, this should be viewed critically, as the reannotation process only considered samples that GPT identified as misclassified. Given the near-deterministic classification, each adjustment results in a predictable improvement in evaluation outcomes for GPT. Although the final decision to adjust labels was made by annotators, independent of GPT’s assessments, the selective review may introduce a bias in favor of GPT."

    GPT-4o Zero-Shot majority predictions determined which 314 samples were flagged for manual reannotation (Section 5.2). In about 64% of reviewed cases the label was changed to GPT's suggested label, so the relabeled test set is partly constructed from GPT-4o's own outputs. Re-running GPT-4o on this set and reporting that its S-score gap over the baseline doubled (0.054 to 0.104, Section 5.3) is therefore a measurement on data selected and partially relabeled by the model being evaluated. The paper itself concedes that each adjustment produces a predictable improvement for GPT. This affects the auxiliary reannotated evaluation, not the headline result on the original 1,592 labels, so the circularity is partial rather than total.

full rationale

The central claim—GPT-4o One-Shot achieving S = 0.7059 versus the HOCON34k baseline S = 0.652 on the original 1,592-sample test set—is not circular. The APIs are external black boxes evaluated against a public dataset, and the baseline is imported from Keller et al. (2024) as a benchmark result, not derived from the present paper's inputs. The one constructive feedback loop is the Section 5 reannotation: GPT-4o's predictions were used to select samples for manual relabeling, and the relabeled data were then used to re-evaluate GPT-4o. The authors explicitly acknowledge this bias and warn that the selective review 'may introduce a bias in favor of GPT.' Because the headline result is based on the original labels and the reannotated result is presented with a candid caveat, the overall circularity score is moderate rather than high. No other self-citation or definitional circularity is load-bearing.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The central claim rests on the accuracy of HOCON34k labels, the comparability of imported baseline numbers, and the temporal stability of proprietary APIs. The reannotation protocol introduces a GPT-selected review loop; the authors acknowledge it may favor GPT-4o.

free parameters (2)
  • Perspective API thresholds = 0.38, 0.5, 0.8
    Probability-to-label mapping thresholds; 0.38 maximizes hate-speech detection in this test, 0.8 yields near-zero recall, so the chosen threshold affects the comparison.
  • HOCON34k baseline threshold = 0.523 (original), 0.557 (reannotated)
    Optimized for the combined S metric by Keller et al. (2024) and re-optimized after reannotation; the central comparison uses the baseline at its best threshold while GPT is not threshold-tuned.
assumptions (5)
  • domain assumption The HOCON34k hate speech guidelines are the correct operational definition for the task
    Used as ground truth for labels and as the prompt for GPT-4o; Sections 1 and 3.
  • domain assumption The original HOCON34k test annotations are accurate enough to rank systems
    All metrics and the baseline threshold depend on these 1,592 labels; Section 5.1 estimates up to 96 may be wrong.
  • domain assumption The baseline results from Keller et al. (2024) are comparable without rerunning
    Section 3.1 states baseline results were taken from the prior paper; no common-run comparison on the original test set.
  • domain assumption Majority vote of three annotators is the correct label for reannotation
    Section 5.2 adopts majority decisions; if annotators are biased, the revised test set inherits the bias.
  • domain assumption Proprietary APIs behave consistently during the test window
    API models are undisclosed and can change; the paper only pins 'June 2024'.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Digital Guardians: Can GPT-4, Perspective API, and Moderation API reliably detect hate speech in reader comments of German online newspapers?." pith.science (2026). https://pith.science/paper/7N6EKHEZ

@misc{pith2026250101256,
  author       = {Pith},
  title        = {Pith review of: Digital Guardians: Can GPT-4, Perspective API, and Moderation API reliably detect hate speech in reader comments of German online newspapers?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7N6EKHEZ}},
  note         = {Machine review of arXiv:2501.01256}
}
read the original abstract

In recent years, toxic content and hate speech have become widespread phenomena on the internet. Moderators of online newspapers and forums are now required, partly due to legal regulations, to carefully review and, if necessary, delete reader comments. This is a labor-intensive process. Some providers of large language models already offer solutions for automated hate speech detection or the identification of toxic content. These include GPT-4o from OpenAI, Jigsaw's (Google) Perspective API, and OpenAI's Moderation API. Based on the selected German test dataset HOCON34k, which was specifically created for developing tools to detect hate speech in reader comments of online newspapers, these solutions are compared with each other and against the HOCON34k baseline. The test dataset contains 1,592 annotated text samples. For GPT-4o, three different promptings are used, employing a Zero-Shot, One-Shot, and Few-Shot approach. The results of the experiments demonstrate that GPT-4o outperforms both the Perspective API and the Moderation API, and exceeds the HOCON34k baseline by approximately 5 percentage points, as measured by a combined metric of MCC and F2-score.

Figures

Figures reproduced from arXiv: 2501.01256 by the authors.

Figure 1
Figure 1. Prompt for GPT-4o Few-Shot Learning: Original Prompt (left) and Translation (right). [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. S-, F1-, and F2-scores of all models on HOCON34k test compared to the reannotated data [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 14 canonical work pages

  1. [1]

    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 islrn pid label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRINGS urlintro eprintur...

  2. [2]

    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 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Fatimah Alkomah and Xiaogang Ma. 2022. https://doi.org/10.3390/info13060273 A Literature Review of Textual Hate Speech Detection Methods and Datasets . Information, 13(6):273

    doi:10.3390/info13060273 2022

  4. [4]

    Branden Chan, Stefan Schweter, and Timo Möller. 2020. http://arxiv.org/abs/2010.10906 German's Next Language Model

    arXiv 2020

  5. [5]

    Davide Chicco and Giuseppe Jurman. 2020. https://doi.org/10.1186/s12864-019-6413-7 The advantages of the Matthews correlation coefficient ( MCC ) over F1 score and accuracy in binary classification evaluation . BMC Genomics, 21(1):6

    doi:10.1186/s12864-019-6413-7 2020

  6. [6]

    Ke-Li Chiu, Annie Collins, and Rohan Alexander. 2021. https://doi.org/10.48550/ARXIV.2103.12407 Detecting Hate Speech with GPT -3

    arXiv doi:10.48550/arxiv.2103.12407 2021

  7. [7]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. http://arxiv.org/abs/1810.04805 BERT : Pre -training of Deep Bidirectional Transformers for Language Understanding

    arXiv 2019

  8. [8]

    ERIC . 2016. https://rm.coe.int/ecri-general-policy/recommendation-no-15-on-combating-hatespeech-germ/16808b5b00 Europäische Kommission gegen Rassismus und Intoleranz Allgemeine Politik - Empfehlung Nr . 15 der ECRI über die Bekämpfung von Hassrede

    work page 2016

Show all 36 references
  1. [9]

    European Parliament . 2022. https://eur-lex.europa.eu/legal-content/DE/TXT/?uri=CELEX

    2022

  2. [10]

    Jonas Glasebach, Max-Emanuel Keller, Alexander Döschl, and Peter Mandl. 2024. https://doi.org/10.1609/icwsm.v18i1.31438 GMHP7k : A Corpus of German Misogynistic Hatespeech Posts . Proceedings of the International AAAI Conference on Web and Social Media, 18:1946--1957

    2024 doi

  3. [11]

    Google Jigsaw . 2024. https://www.perspectiveapi.com/ Perspective API . [Accessed: 2024-09-21]

    2024

  4. [12]

    Keyan Guo, Alexander Hu, Jaden Mu, Ziheng Shi, Ziming Zhao, Nishant Vishwamitra, and Hongxin Hu. 2024. https://doi.org/10.48550/ARXIV.2401.03346 An Investigation of Large Language Models for Real - World Hate Speech Detection

    2024 doi arXiv

  5. [13]

    Hossein Hosseini, Sreeram Kannan, Baosen Zhang, and Radha Poovendran. 2017. https://doi.org/10.48550/ARXIV.1702.08138 Deceiving Google 's Perspective API Built for Detecting Toxic Comments

    2017 doi arXiv

  6. [14]

    Othman Istaiteh, Razan Al-Omoush, and Sara Tedmori. 2020. https://doi.org/10.1109/IDSTA50958.2020.9264052 Racist and Sexist Hate Speech Detection : Literature Review . In 2020 International Conference on Intelligent Data Science Technologies and Applications ( IDSTA ) , pages 95--99

    2020

  7. [15]

    Md Saroar Jahan and Mourad Oussalah. 2023. https://doi.org/10.1016/j.neucom.2023.126232 A systematic review of hate speech automatic detection using natural language processing . Neurocomputing, 546:126232

    2023

  8. [16]

    Md Saroar Jahan, Mourad Oussalah, Djamila Romaissa Beddia, Jhuma kabir Mim, and Nabil Arhab. 2024. http://arxiv.org/abs/2404.00303 A Comprehensive Study on NLP Data Augmentation for Hate Speech Detection : Legacy Methods , BERT , and LLMs

    2024 arXiv

  9. [17]

    Sylvia Jaki and Stefan Steiger. 2023. https://doi.org/10.1007/978-3-662-65964-9_1 Hate Speech online: Hartknäckiges Phänomen und interdisziplinärer Forschungsgegenstand . In Sylvia Jaki and Stefan Steiger, editors, Digitale Hate Speech , pages 1--14. Springer Berlin Heidelberg

    2023 doi

  10. [18]

    Johnson and Taghi M

    Justin M. Johnson and Taghi M. Khoshgoftaar. 2019. https://doi.org/10.1186/s40537-019-0192-5 Survey on deep learning with class imbalance . Journal of Big Data, 6(1):27

    2019 doi

  11. [19]

    Max-Emanuel Keller, Maximilian Auch, Alexander Döschl, Fabian Vlk, Julian Quernheim, Mike Hartmann, Peter Mandl, Alexander Kaul, and Markus Franz. 2024. https://doi.org/10.1007/978-3-031-78090-5_18 HOCON34k : A Corpus of Hate speech in Online Comments from German Newspapers . ...

    2024 doi

  12. [20]

    Tran, Yi Tay, Jeffrey Sorensen, Jai Gupta, Donald Metzler, and Lucy Vasserman

    Alyssa Lees, Vinh Q. Tran, Yi Tay, Jeffrey Sorensen, Jai Gupta, Donald Metzler, and Lucy Vasserman. 2022. https://doi.org/10.48550/ARXIV.2202.11176 A New Generation of Perspective API : Efficient Multilingual Character -level Transformers

    2022 doi arXiv

  13. [21]

    Lingyao Li, Lizhou Fan, Shubham Atreja, and Libby Hemphill. 2023. https://doi.org/10.48550/ARXIV.2304.10619 " HOT " ChatGPT : The promise of ChatGPT in detecting and discriminating hateful, offensive, and toxic comments on social media

    2023 doi arXiv

  14. [22]

    Hiren Madhu, Shrey Satapara, Sandip Modha, Thomas Mandl, and Prasenjit Majumder. 2023. https://doi.org/10.1016/j.eswa.2022.119342 Detecting offensive speech in conversational code-mixed dialogue on social media: A contextual dataset and benchmark experiments . Expert Systems w...

    2023

  15. [23]

    Thomas Mandl. 2023. https://doi.org/10.1007/978-3-662-65964-9_6 KI - Verfahren für die Hate Speech Erkennung : Die Gestaltung von Ressourcen für das maschinelle Lernen und ihre Zuverlässigkeit . In Sylvia Jaki and Stefan Steiger, editors, Digitale Hate Speech , pages 111--130....

    2023 doi

  16. [24]

    Todor Markov, Chong Zhang, Sandhini Agarwal, Tyna Eloundou, Teddy Lee, Steven Adler, Angela Jiang, and Lilian Weng. 2022. https://doi.org/10.48550/ARXIV.2208.03274 A Holistic Approach to Undesired Content Detection in the Real World

    2022 doi arXiv

  17. [25]

    Daniel Matter, Miriam Schirmer, Nir Grinberg, and Jürgen Pfeffer. 2024. https://doi.org/10.48550/ARXIV.2401.02001 Close to Human - Level Agreement : Tracing Journeys of Violent Speech in Incel Posts with GPT -4- Enhanced Annotations

    2024 doi arXiv

  18. [26]

    Meta . 2024. https://transparency.meta.com/de-de/policies/community-standards/hate-speech/ Hassrede Transparency Center . [Accessed: 2024-09-21]

    2024

  19. [27]

    Gianluca Nogara, Francesco Pierri, Stefano Cresci, Luca Luceri, Petter Törnberg, and Silvia Giordano. 2023. https://doi.org/10.48550/ARXIV.2312.12651 Toxic Bias : Perspective API Misreads German as More Toxic

    2023 doi arXiv

  20. [28]

    OpenAI . 2023. https://doi.org/10.48550/ARXIV.2303.08774 GPT -4 Technical Report . [Accessed: 2024-09-21]

    2023 doi arXiv

  21. [29]

    OpenAI . 2024. https://platform.openai.com OpenAI Platform . [Accessed: 2024-09-21]

    2024

  22. [30]

    Ronghao Pan, José Antonio García-Díaz, and Rafael Valencia-García. 2024. https://doi.org/10.32604/cmes.2024.049631 Comparing Fine - Tuning , Zero and Few - Shot Strategies with Large Language Models in Hate Speech Detection in English . Computer Modeling in Engineering & Scien...

    2024

  23. [31]

    Perspective . 2024. https://developers.perspectiveapi.com/s/?language=en_US Perspective Developers . [Accessed: 2024-09-21]

    2024

  24. [32]

    David MW Powers. 2020. https://arxiv.org/abs/2010.16061 Evaluation: from precision, recall and F -measure to ROC , informedness, markedness and correlation . arXiv:2010.16061

    2020 arXiv

  25. [33]

    Anchal Rawat, Santosh Kumar, and Surender Singh Samant. 2024. https://doi.org/10.1002/wics.1648 Hate speech detection in social media: Techniques , recent trends, and future challenges . WIREs Computational Statistics, 16(2):e1648

    2024 doi

  26. [34]

    Tran, Sebastian Ruder, Jai Gupta, Hyung Won Chung, Dara Bahri, Zhen Qin, Simon Baumgartner, Cong Yu, and Donald Metzler

    Yi Tay, Vinh Q. Tran, Sebastian Ruder, Jai Gupta, Hyung Won Chung, Dara Bahri, Zhen Qin, Simon Baumgartner, Cong Yu, and Donald Metzler. 2021. https://doi.org/10.48550/ARXIV.2106.12672 Charformer: Fast Character Transformers via Gradient -based Subword Tokenization

    2021 doi arXiv

  27. [35]

    Collins Udanor and Chinatu C. Anyanwu. 2019. https://doi.org/10.1108/DTA-01-2019-0007 Combating the challenges of social media hate speech in a polarized society: A Twitter ego lexalytics approach . Data Technologies and Applications, 53(4):501--527

    2019 doi

  28. [36]

    Gomez, Lukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. 2017. http://arxiv.org/abs/1706.03762 Attention Is All You Need

    2017 arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.