Pith. sign in

REVIEW 3 major objections 5 minor 71 references

Making FETCH! Happen: Finding Emergent Dog Whistles Through Common Habitats

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper introduces FETCH!, a benchmark for discovering novel dog whistles in social media, and EarShot, a baseline the authors claim raises $F_{0.5}$ by 2-20 points over prior methods.

desk verdict Useful benchmark, honest limitations, but the headline F-scores measure retrieval from a closed lexicon, not discovery of novel dog whistles. read the letter →

arxiv 2412.12072 v2 pith:MOY2TMTY submitted 2024-12-16 cs.CL

Kuleen Sasse , Carlos Aguirre , Isabel Cachola , Sharon Levy , Mark Dredze This is my paper · ORCID
classification cs.CL
keywords dogwhistlescodedlanguagehatespeechdetectioneuphemismsocialmedialargemodelssentenceembeddingsbenchmark
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

Dog whistles are coded phrases that say something innocuous to most readers while signaling a second, often hateful message to an in-group, which lets them slip past moderation filters. The paper tries to establish that this problem can be tackled as a discovery task: given a corpus and a small set of known seed dog whistles, a system should surface other dog whistles that are not in any lexicon. To that end it introduces FETCH!, a benchmark with three social-media case studies (dense synthetic Reddit posts, high-prevalence Gab, sparse Twitter), and EarShot, a baseline that retrieves posts similar to seed-whistle posts and then filters and extracts candidate terms. The paper reports that all three existing methods perform poorly on FETCH!, with the best prior $F_{0.5}$ at 5.91 on Synthetic and below 3 on the other two, while EarShot's best configurations reach 14.55, 5.70, and 4.63 on Synthetic, Balanced, and Realistic. A sympathetic reader would care because curated dog-whistle lexicons are slow to update, and an automated discovery pipeline could help moderators and researchers track coded hate speech as it emerges.

What carries the argument

FETCH! is the evaluation object: a corpus plus seed dog whistles with a stratified train/test split and Data Potential Recall as the denominator. EarShot is the mechanism that carries the argument. Its first stage maps each post to a vector using a sentence encoder, so posts similar in meaning are close together, and stores these vectors in a vector database. For every post that contains a seed whistle, the system retrieves its nearest neighbors from the database, capturing posts that share meaning, sentiment, or intent without sharing exact words. In the PREDICT path, those neighbors are filtered by a hate/toxicity classifier or an LLM yes/no prompt, then passed to a keyword extractor that returns top-k candidate terms; in the DIRECT path, an LLM reads the neighbor posts and returns dog whistles in a structured list. The load-bearing idea is that novel whistles live in the semantic neighborhood of known whistles, not in token overlap.

What would settle it

Have independent human annotators label a random sample of FETCH! positive posts as genuine dog whistles or not, then recompute EarShot's precision and $F_{0.5}$ against the corrected labels; if precision drops substantially on the Balanced or Realistic sets, the reported 2-20 point gains are partly artifacts of noisy automatic ground truth.

Watch

Extended reading notes

Core claim

The paper's central claim is that dog whistle discovery can be formalized and measurably improved. FETCH! splits known dog whistles into seed and held-out sets, then asks a system to find the held-out whistles in a corpus, with recall normalized by which whistles actually occur in that corpus (Data Potential Recall). On this benchmark, Word2Vec/Phrase2Vec, masked-language-model fill-in, and euphemistic-phrase detectors all perform poorly: their best $F_{0.5}$ is 5.91 on Synthetic and 2.47 or lower on the other two, which the paper reads as evidence that lexicon-based and token-filling approaches do not transfer to discovery. EarShot, by contrast, embeds every post, retrieves nearest neighbors of posts containing seed whistles, and either asks an LLM directly for the whistles or filters candidates with a hate/toxicity classifier or LLM and then extracts keywords; the best configurations reach $F_{0.5}$ of 14.55 on Synthetic, 5.70 on Balanced, and 4.63 on Realistic. The paper presents EarShot not as a finished detector but as a strong baseline that outperforms prior methods by 2 to 20 F-points while leaving subtle, recent, and emoji-based whistles largely undiscovered.

Load-bearing premise

The benchmark's ground truth is produced automatically from an existing word list and from AI annotations, with no human-verified labels, so if those automatic labels are wrong every score in the paper is mismeasured.

Editorial extensions

If this is right

  • Any future discovery method can be tested on the same three habitats with the same seed/holdout split, so results across systems are directly comparable.
  • Lexicon-based euphemism detectors such as Word2Vec, masked language models, and phrase detectors perform poorly on FETCH!, with best $F_{0.5}$ at 5.91 on Synthetic and below 3 on Balanced and Realistic, so they are unlikely to work as drop-in discovery tools.
  • EarShot's best configurations favor precision over recall, so a human reviewer sees relatively few false candidates, but the tradeoff is that explicit slurs are found while subtle and recent phrases are missed.
  • Direct LLM prompting recovers far more true dog whistles (higher Data Potential Recall) at much lower precision, and in the sparse Realistic corpus the best DIRECT run scores 0 in $F_{0.5}$.
  • Because recall is computed against whistles that actually appear in each corpus, a method cannot gain credit by listing known terms that are absent, so lexicon coverage alone will not inflate scores.

Reading between the lines

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

  • One implication the paper leaves implicit is that its automatically generated labels make the reported scores an upper-bound estimate; a human-adjudicated version of FETCH! would likely shift both precision and recall.
  • A natural next experiment, not run in the paper, is to apply EarShot longitudinally on time-sliced corpora and check whether its retrievals predate a whistle's appearance in curated lexicons, which would directly test emergent discovery.
  • The bias toward overt slurs under hate-speech filtering suggests that a filter-free variant ranking neighbors by embedding distance alone could recover recency-dependent whistles like 'jogger' or 'vegan cat' that the paper reports missing.
  • The same retrieval-plus-filter architecture could transfer to other coded-language domains with a seed vocabulary, such as drug slang or underground-economy jargon, because it does not depend on a particular lexicon.
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

3 major / 5 minor

Summary. This paper introduces FETCH!, a benchmark for discovering novel dog whistles in social media corpora, with three scenarios: Synthetic (Reddit/Silent Signals), Balanced (Gab), and Realistic (Twitter). Given a corpus and seed dog whistles, systems must propose additional dog whistles. The authors evaluate Word2Vec/Phrase2Vec, MLM fill-in-the-blank methods, EPD, and their EarShot pipeline, which combines sentence embeddings, nearest-neighbor retrieval, LLM/BERT filtering, and keyword extraction or direct LLM prompting. They report that EarShot outperforms the baselines by 2 to 20 F0.5 points, with best scores of 14.55 on Synthetic, 5.70 on Balanced, and 4.63 on Realistic. The paper also includes a manual review of novel dog whistles in Appendix G.

Significance. The task is well motivated and the resource could be useful to the community. EarShot is described in enough detail to be reimplemented, code is released, and the use of Data Potential Recall is a thoughtful adjustment for sparse positive settings. However, the evaluation design measures retrieval of known lexicon items rather than discovery of novel dog whistles, because the positive labels are produced by regular-expression matching of the Mendelsohn et al. (2023) lexicon or by GPT-4 annotations, and scoring requires exact matches to known surface forms. This construct-validity gap is acknowledged in Section 9 but not resolved, and it affects every headline number in the paper. If the paper is reframed as a retrieval baseline over known dog whistles and supplemented with a human-validated evaluation of novel discoveries, it would be a solid contribution.

major comments (3)
  1. [Sections 3.1.3, 3.3, 9] The benchmark cannot credit out-of-lexicon discoveries. In the Balanced and Realistic scenarios, ground truth is generated by regular-expression matching of the Mendelsohn et al. (2023) lexicon (Section 3.1.3), and in the Synthetic scenario by GPT-4 annotations in Silent Signals (Section 3.1.1). Section 3.3 then scores a prediction only if it exactly matches a root or surface form. As a result, any prediction outside the curated lexicon is counted as a false positive, and the F0.5 numbers in Tables 2-5 measure retrieval of known dog whistles rather than discovery of emergent ones. The manual review in Appendix G is the only evidence of novel discovery, but it is not scored, not systematic, and in the Realistic scenario it found no new dog whistles (Appendix G.3). Section 9 concedes the lack of human-labeled corpora, but this is not only a label-noise problem: the evaluation design makes novel true positives impossible. I recommend adding a held-out human-annotated set of genuinely new dog whistles, or making manually verified discoveries a primary reported metric.
  2. [Tables 3-5 and Appendices D-F] The reported best F0.5 scores are selected by sweeping prediction thresholds (k = 50 to 25,600 for MLM/EPD and EarShot) and taking the maximum on the test set, with no development split or a priori threshold rule. This makes the headline 2-to-20-point improvements optimistic and difficult to reproduce. The authors should either fix thresholds before test evaluation, report development-set selection, or clearly label the selected maxima as upper bounds. This issue is load-bearing because the superiority claim rests on these selected values.
  3. [Abstract and Section 4.1] The abstract's claim that state-of-the-art systems fail to achieve meaningful results is stronger than the evaluation supports. The compared methods are Word2Vec/Phrase2Vec, MLM, and EPD, which are general embedding and euphemism-detection baselines; the most relevant prior dog-whistle systems (Mendelsohn et al. 2023 and Kruk et al. 2024, discussed in Section 2.4) are not evaluated as systems. The claim should be narrowed to the tested methods, or those prior systems should be included in the comparison.
minor comments (5)
  1. [Abstract/Table 2] The abstract and introduction say state-of-the-art systems achieve less than 5% F-score, but the unigram Word2Vec model on Synthetic in Table 2 reaches F0.5 = 5.91; the claim should be qualified to the specific baselines or specific scenarios.
  2. [Section 6] The first sentence of Section 6 contains a grammatical error ('We differences between models'), and there are several other typos, including 'phenomenia' in Section 2.3 and 'an modification' in Appendix G.2.
  3. [Section 3.2] The final sentence of Section 3.2 is garbled and should be split into two sentences to explain what metadata is used for stratified splitting and why the Synthetic scenario is handled differently.
  4. [Appendix G.3] The statement that the best Realistic system found no new dog whistles should be acknowledged in the main discussion, because it substantially qualifies the discovery claim for the realistic setting.
  5. [Section 4.1.4 and Section 3.1.2] The paper does not report the exact number of Gab posts collected after scraping or the composition of the corpus after filtering; additional data-collection details would improve reproducibility.

Circularity Check

1 steps flagged · score 6.0 of 10

FETCH! defines 'new dog whistles' as held-out entries of the same Mendelsohn lexicon used to generate its regex labels, so EarShot's reported F-scores measure closed-lexicon retrieval rather than novel discovery; the discovery claim reduces by construction to retrieval of known terms.

  1. self definitional [Section 3.1.3 (Realistic Twitter) and Section 3.3 (Metrics for Evaluation); task definition and seed split in Section 3.2]
    ""1 million tweets containing dog whistles identified using regular expression from a set of ground truth dog whistles"; "We use the dog whistle dataset released by Mendelsohn et al. (2023) as our ground truth"; "all metrics count a positive match for a single root dog whistle as returning an exact match of one of the surface forms or the root itself.""

    FETCH! claims to evaluate discovery of 'emerging or previously unknown dog whistles,' but the Balanced and Realistic positive labels are regex matches from the Mendelsohn et al. lexicon, and the metrics credit only exact matches to surface forms or roots from that same lexicon. Thus 'new dog whistle' is operationally identical to 'held-out member of the known lexicon': any prediction outside the lexicon is a false positive, and no genuinely novel term can appear in the numerator of precision, DPR, or F0.5. EarShot's headline scores (14.55 Synthetic, 5.70 Balanced, 4.63 Realistic) are therefore retrieval scores over a closed list, not measurements of novel discovery.

full rationale

There is no equation-level circularity in EarShot itself: the pipeline of sentence embeddings, nearest-neighbor retrieval, LLM/BERT filtering, and keyword extraction is a legitimate baseline architecture, and its components are externally sourced. The self-citations to Bernice and related prior work are not load-bearing for the main claim. The central circularity is in the benchmark's operationalization of 'novel dog whistle.' The paper splits a known lexicon into seed and test portions, labels the test portion via regex or GPT-4 annotations, and then scores predictions by exact match against that same fixed lexicon. Because any output that is not a surface form or root in the lexicon is scored as a false positive, the benchmark cannot represent the discovery of a genuinely new dog whistle. Consequently, the reported F-score improvements support the claim that EarShot retrieves held-out entries of a known dog-whistle list, not that it discovers emergent dog whistles. The paper's Appendix G manual review does provide some independent evidence of novel terms, but it is presented qualitatively and is not part of the quantitative precision/DPR/F0.5 results. Section 9 openly acknowledges the lack of dedicated human-labeled corpora and the risk of LLM pretraining contamination, which is honest but does not repair the definitional gap between the task's stated goal and what the metric actually rewards. Overall, this is a partial but central circularity: the headline 'novel discovery' result reduces by construction to closed-lexicon retrieval, while the underlying system remains a reasonable baseline for the narrower retrieval task.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The benchmark's validity rests on external lexicons and automated labels rather than on fitted model parameters. The only fitted quantities are evaluation-side choices (thresholds, n-gram ranges) selected on the test set. No new physical or theoretical entities are introduced.

free parameters (3)
  • prediction threshold k = varies per model and dataset, e.g., 50, 400, 6400
    MLM, EPD, and EarShot results are reported at the threshold that maximizes F0.5 on the test set for each scenario; no held-out validation split is used, so reported scores are test-optimized.
  • keyword extraction n-gram range = 1-gram, 1-2 grams, or 1-3 grams, chosen per dataset
    For EarShot PREDICT, RAKE and YAKE require an n-gram range; the range is selected by best F0.5 on the test set, which inflates apparent performance.
  • number of nearest neighbors = 1
    EarShot retrieves the single closest post vector to each seed post; this design choice is not varied or ablated.
assumptions (5)
  • domain assumption The Mendelsohn et al. (2023) lexicon of about 340 dog whistle roots and surface forms is accurate and complete enough to define ground truth for Balanced and Realistic scenarios.
    Used as seed and reference labels in Section 3.2; if a dog whistle is missing from the lexicon, recall cannot be measured correctly.
  • domain assumption Regular-expression matching of lexicon surface forms finds all and only dog whistle occurrences in the Gab and Twitter corpora.
    Realistic and Balanced labels are regex-generated, not human-verified; Section 3.1.3 and Section 9 acknowledge false positives.
  • domain assumption Silent Signals (Kruk et al., 2024) GPT-4-identified dog whistle labels are reliable enough for the Synthetic scenario.
    The Synthetic scenario uses an LLM-annotated dataset as ground truth; Section 3.1.1.
  • domain assumption EarShot LLM outputs are not substantially inflated by pretraining memorization of the dog whistle lexicon.
    Authors state in Section 9 that they cannot ensure the models were not exposed to the evaluation examples during pretraining; if memorization occurred, discovery scores are overstated.
  • domain assumption Nearest-neighbor semantic similarity to seed dog whistles is a valid discovery signal.
    EarShot's retrieval stage assumes embedding closeness to seed posts surfaces posts that contain new dog whistles; no separate validation of this assumption is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Making FETCH! Happen: Finding Emergent Dog Whistles Through Common Habitats." pith.science (2026). https://pith.science/paper/MOY2TMTY

@misc{pith2026241212072,
  author       = {Pith},
  title        = {Pith review of: Making FETCH! Happen: Finding Emergent Dog Whistles Through Common Habitats},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MOY2TMTY}},
  note         = {Machine review of arXiv:2412.12072}
}
read the original abstract

WARNING: This paper contains content that maybe upsetting or offensive to some readers. Dog whistles are coded expressions with dual meanings: one intended for the general public (outgroup) and another that conveys a specific message to an intended audience (ingroup). Often, these expressions are used to convey controversial political opinions while maintaining plausible deniability and slip by content moderation filters. Identification of dog whistles relies on curated lexicons, which have trouble keeping up to date. We introduce FETCH!, a task for finding novel dog whistles in massive social media corpora. We find that state-of-the-art systems fail to achieve meaningful results across three distinct social media case studies. We present EarShot, a strong baseline system that combines the strengths of vector databases and Large Language Models (LLMs) to efficiently and effectively identify new dog whistles.

Figures

Figures reproduced from arXiv: 2412.12072 by the authors.

Figure 1
Figure 1. Comparison between dog whistle detection [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Flowchart for the EarShot System. Two paths can be taken after doing the nearest neighbor lookup. Path [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Word2Vec/Phrase2Vec F0.5 performance vs the number of words/phrases returned by the model. Plot is on log scale [PITH_FULL_IMAGE:figures/full_fig_p018_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: MLM and EPD F0.5 performance vs the prediction threshold. Plot is on log scale [PITH_FULL_IMAGE:figures/full_fig_p019_4.png]
Figure 5
Figure 5. Figure 5: F0.5 averaged across BERT based filtering methods displaying average performance across keyword extraction models for EarShot BERT PREDICT vs prediction threshold across all three datasets. Plot is on log scale [PITH_FULL_IMAGE:figures/full_fig_p020_5.png]
Figure 6
Figure 6. Figure 6: F0.5 averaged across keyword extraction models displaying average performance across BERT based filtering models for EarShot BERT PREDICT vs prediction threshold across all three datasets. Plot is on log scale [PITH_FULL_IMAGE:figures/full_fig_p020_6.png]
Figure 7
Figure 7. Figure 7: F0.5 averaged across LLM based filtering methods displaying average performance across keyword extraction models for EarShot LLM PREDICT vs prediction threshold across all three datasets. Plot is on log scale [PITH_FULL_IMAGE:figures/full_fig_p021_7.png]
Figure 8
Figure 8. Figure 8: F0.5 averaged across keyword extraction models displaying average performance across LLM based filtering models for EarShot LLM PREDICT vs prediction threshold across all three datasets. Plot is on log scale. G Novel Dog Whistles We reviewed the top systems’ output to …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

71 extracted references · 38 canonical work pages

  1. [1]

    online" 'onlinestring :=

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

  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]

    Mathilda kerlund. 2022. Dog whistling far-right code words: the case of ‘culture enricher'on the swedish web. Information, Communication & Society, 25(12):1808--1825

    work page 2022

  4. [4]

    Bethany L Albertson. 2015. Dog-whistle politics: Multivocal communication and religious appeals. Political Behavior, 37:3--26

    work page 2015

  5. [5]

    Tatsuya Aoki, Ryohei Sasano, Hiroya Takamura, and Manabu Okumura. 2017. https://doi.org/10.18653/v1/D17-1246 Distinguishing J apanese non-standard usages from standard ones . In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 2323--2328, Copenhagen, Denmark. Association for Computational Linguistics

    work page doi:10.18653/v1/d17-1246 2017

  6. [6]

    Prashanth Bhat and Ofra Klein. 2020. Covert hate speech: White nationalists and dog whistle communication on twitter. Twitter, the public sphere, and the chaos of online deliberation, pages 151--172

    work page 2020

  7. [7]

    Steven Bird and Edward Loper. 2004. https://aclanthology.org/P04-3031 NLTK : The natural language toolkit . In Proceedings of the ACL Interactive Poster and Demonstration Sessions , pages 214--217, Barcelona, Spain. Association for Computational Linguistics

    2004

  8. [8]

    o rn R \

    Max Boholm, Bj \"o rn R \"o nnerstrand, Ellen Breitholtz, Robin Cooper, Elina Lindgren, Gregor Rettenegger, and Asad Sayeed. 2024. https://doi.org/10.18653/v1/2024.lchange-1.14 Can political dogwhistles be predicted by distributional methods for analysis of lexical semantic change? In Proceedings of the 5th Workshop on Computational Approaches to Historic...

    work page doi:10.18653/v1/2024.lchange-1.14 2024

Show all 71 references
  1. [9]

    Max Boholm and Asad Sayeed. 2023. https://doi.org/10.18653/v1/2023.lchange-1.6 Political dogwhistles and community divergence in semantic change . In Proceedings of the 4th Workshop on Computational Approaches to Historical Language Change, pages 53--65, Singapore. Association...

    2023 doi

  2. [10]

    Luke Breitfeller, Emily Ahn, David Jurgens, and Yulia Tsvetkov. 2019. https://doi.org/10.18653/v1/D19-1176 Finding microaggressions in the wild: A case for locating elusive phenomena in social media posts . In Proceedings of the 2019 Conference on Empirical Methods in Natural ...

    2019 doi

  3. [11]

    Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffr...

    2020 arXiv

  4. [12]

    Ricardo Campos, V \' tor Mangaravite, Arian Pasquali, Al \' pio M \'a rio Jorge, C \'e lia Nunes, and Adam Jatowt. 2018. A text feature based automatic keyword extraction method for single documents. In European conference on information retrieval, pages 684--691. Springer

    2018

  5. [13]

    Chroma. 2024. chroma. https://github.com/chroma-core/chroma

    2024

  6. [14]

    Alexandra DeLucia. 2020. Little bird. https://github.com/aadelucia/littlebird

    2020

  7. [15]

    Alexandra DeLucia, Shijie Wu, Aaron Mueller, Carlos Aguirre, Philip Resnik, and Mark Dredze. 2022. https://doi.org/10.18653/v1/2022.emnlp-main.415 Bernice: A multilingual pre-trained encoder for T witter . In Proceedings of the 2022 Conference on Empirical Methods in Natural L...

    2022 doi

  8. [16]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. https://arxiv.org/abs/1810.04805 Bert: Pre-training of deep bidirectional transformers for language understanding . Preprint, arXiv:1810.04805

    2019 arXiv

  9. [17]

    Kevin Drakulich, Kevin H Wozniak, John Hagan, and Devon Johnson. 2020. Race and policing in the 2016 presidential election: Black lives matter, the police, and dog whistle politics. Criminology, 58(2):370--402

    2020

  10. [18]

    Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, Anirudh Goyal, Anthony Hartshorn, Aobo Yang, Archi Mitra, Archie Sravankumar, Artem Korenev, Arthur Hinsvark, Arun Rao, Aston Zha...

    2024 arXiv

  11. [19]

    Mai ElSherief, Caleb Ziems, David Muchlinski, Vaishnavi Anupindi, Jordyn Seybolt, Munmun De Choudhury, and Diyi Yang. 2021. https://doi.org/10.18653/v1/2021.emnlp-main.29 Latent hatred: A benchmark for understanding implicit hate speech . In Proceedings of the 2021 Conference ...

    2021 doi

  12. [20]

    Lei Gao, Alexis Kuppersmith, and Ruihong Huang. 2017. https://aclanthology.org/I17-1078 Recognizing explicit and implicit hate speech using a weakly supervised two-path bootstrapping approach . In Proceedings of the Eighth International Joint Conference on Natural Language Pro...

    2017

  13. [21]

    Robert E Goodin and Michael Saward. 2005. Dog whistles and democratic mandates. The Political Quarterly, 76(4):471--476

    2005

  14. [22]

    Maarten Grootendorst. 2020. https://doi.org/10.5281/zenodo.4461265 Keybert: Minimal keyword extraction with bert

    2020 doi

  15. [23]

    Takuro HADA, Yuichi SEI, Yasuyuki TAHARA, and Akihiko OHSUGA. 2020. https://doi.org/10.1109/iCCECE49321.2020.9231109 Codewords detection in microblogs focusing on differences in word use between two corpora . In 2020 International Conference on Computing, Electronics & Communi...

    2020

  16. [24]

    Ian Haney-L \'o pez. 2014. Dog whistle politics: How coded racial appeals have reinvented racism and wrecked the middle class. Oxford University Press

    2014

  17. [25]

    Thomas Hartvigsen, Saadia Gabriel, Hamid Palangi, Maarten Sap, Dipankar Ray, and Ece Kamar. 2022. https://arxiv.org/abs/2203.09509 Toxigen: A large-scale machine-generated dataset for adversarial and implicit hate speech detection . Preprint, arXiv:2203.09509

    2022 arXiv

  18. [26]

    Robert Henderson and Elin McCready. 2018. How dogwhistles work. In New Frontiers in Artificial Intelligence: JSAI-isAI Workshops, JURISIN, SKL, AI-Biz, LENLS, AAA, SCIDOCA, kNeXI, Tsukuba, Tokyo, November 13-15, 2017, Revised Selected Papers 9, pages 231--240. Springer

    2018

  19. [27]

    Robert Henderson and Elin McCready. 2020. https://aclanthology.org/2020.pam-1.10 Towards functional, agent-based models of dogwhistle communication . In Proceedings of the Probability and Meaning Conference (PaM 2020), pages 73--77, Gothenburg. Association for Computational Li...

    2020

  20. [28]

    o rn R \

    Niclas Hertzberg, Robin Cooper, Elina Lindgren, Bj \"o rn R \"o nnerstrand, Gregor Rettenegger, Ellen Breitholtz, and Asad Sayeed. 2022. https://doi.org/10.18653/v1/2022.woah-1.16 Distributional properties of political dogwhistle representations in S wedish BERT . In Proceedin...

    2022 doi

  21. [29]

    Jon Hurwitz and Mark Peffley. 2005. Playing the race card in the post--willie horton era: The impact of racialized code words on support for punitive crime policy. Public Opinion Quarterly, 69(1):99--112

    2005

  22. [30]

    Albert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, Lélio Renard Lavaud, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas ...

    2023 arXiv

  23. [31]

    Weld, Luke Zettlemoyer, and Omer Levy

    Mandar Joshi, Danqi Chen, Yinhan Liu, Daniel S. Weld, Luke Zettlemoyer, and Omer Levy. 2020. https://arxiv.org/abs/1907.10529 Spanbert: Improving pre-training by representing and predicting spans . Preprint, arXiv:1907.10529

    2020 arXiv

  24. [32]

    Sedrick Scott Keh. 2022. https://doi.org/10.18653/v1/2022.flp-1.24 Exploring euphemism detection in few-shot and zero-shot settings . In Proceedings of the 3rd Workshop on Figurative Language Processing (FLP), pages 167--172, Abu Dhabi, United Arab Emirates (Hybrid). Associati...

    2022 doi

  25. [33]

    Justin Khoo. 2017. https://www.jstor.org/stable/26529437 Code words in political discourse . Philosophical Topics, 45(2):33--64

    2017

  26. [34]

    Taehoon Kim and Kevin Wurster. 2024. emoji. https://github.com/carpedm20/emoji/

    2024

  27. [35]

    Julia Kruk, Michela Marchini, Rijul Magu, Caleb Ziems, David Muchlinski, and Diyi Yang. 2024. https://doi.org/10.18653/v1/2024.acl-long.675 Silent signals, loud impact: LLM s for word-sense disambiguation of coded dog whistles . In Proceedings of the 62nd Annual Meeting of the...

    2024 doi

  28. [36]

    Christopher N Lasch. 2016. Sanctuary cities and dog-whistle politics. New Eng. J. on Crim. & Civ. Confinement, 42:159

    2016

  29. [37]

    Patrick Lee, Anna Feldman, and Jing Peng. 2022. https://doi.org/10.18653/v1/2022.flp-1.27 A report on the euphemisms detection shared task . In Proceedings of the 3rd Workshop on Figurative Language Processing (FLP), pages 184--190, Abu Dhabi, United Arab Emirates (Hybrid). As...

    2022 doi

  30. [38]

    Omer Levy and Yoav Goldberg. 2014. https://doi.org/10.3115/v1/P14-2050 Dependency-based word embeddings . In Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 302--308, Baltimore, Maryland. Association for C...

    2014 doi

  31. [39]

    Rijul Magu and Jiebo Luo. 2018. https://doi.org/10.18653/v1/W18-5112 Determining code words in euphemistic hate speech using word embedding networks . In Proceedings of the 2nd Workshop on Abusive Language Online ( ALW 2) , pages 93--100, Brussels, Belgium. Association for Com...

    2018 doi

  32. [40]

    Abulimiti Maimaitituoheti, Yang Yong, and Fan Xiaochao. 2022. https://doi.org/10.18653/v1/2022.flp-1.2 A prompt based approach for euphemism detection . In Proceedings of the 3rd Workshop on Figurative Language Processing (FLP), pages 8--12, Abu Dhabi, United Arab Emirates (Hy...

    2022 doi

  33. [41]

    Binny Mathew, Punyajoy Saha, Seid Muhie Yimam, Chris Biemann, Pawan Goyal, and Animesh Mukherjee. 2022. https://arxiv.org/abs/2012.10289 Hatexplain: A benchmark dataset for explainable hate speech detection . Preprint, arXiv:2012.10289

    2022 arXiv

  34. [42]

    Tali Mendelberg. 2001. The Race Card: Campaign Strategy, Implicit Messages, and the Norm of Equality. Princeton University Press

    2001

  35. [43]

    Julia Mendelsohn, Ronan Le Bras, Yejin Choi, and Maarten Sap. 2023. https://doi.org/10.18653/v1/2023.acl-long.845 From dogwhistles to bullhorns: Unveiling coded rhetoric with language models . In Proceedings of the 61st Annual Meeting of the Association for Computational Lingu...

    2023 doi

  36. [44]

    Rada Mihalcea and Paul Tarau. 2004. https://aclanthology.org/W04-3252 T ext R ank: Bringing order into text . In Proceedings of the 2004 Conference on Empirical Methods in Natural Language Processing, pages 404--411, Barcelona, Spain. Association for Computational Linguistics

    2004

  37. [45]

    Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. https://arxiv.org/abs/1301.3781 Efficient estimation of word representations in vector space . Preprint, arXiv:1301.3781

    2013 arXiv

  38. [46]

    Dat Quoc Nguyen, Thanh Vu, and Anh Tuan Nguyen. 2020. https://doi.org/10.18653/v1/2020.emnlp-demos.2 BERT weet: A pre-trained language model for E nglish tweets . In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, ...

    2020 doi

  39. [47]

    Nicol \'a s Benjam \' n Ocampo, Ekaterina Sviridova, Elena Cabrio, and Serena Villata. 2023. https://doi.org/10.18653/v1/2023.eacl-main.147 An in-depth analysis of implicit and subtle hate speech messages . In Proceedings of the 17th Conference of the European Chapter of the A...

    2023 doi

  40. [48]

    OpenAI, Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, Red Avila, Igor Babuschkin, Suchir Balaji, Valerie Balcom, Paul Baltescu, Haiming Bao, Mohammad Bavarian, Jeff ...

    2024 arXiv

  41. [49]

    Jing Qian, Mai ElSherief, Elizabeth Belding, and William Yang Wang. 2019. https://doi.org/10.18653/v1/N19-1305 Learning to decipher hate symbols . In Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Langu...

    2019 doi

  42. [50]

    Anne Quaranto. 2022. Dog whistles, covertly coded speech, and the practices that enable them. Synthese, 200(4):330

    2022

  43. [51]

    Radim Rehurek and Petr Sojka. 2011. Gensim--python framework for vector space modelling. NLP Centre, Faculty of Informatics, Masaryk University, Brno, Czech Republic, 3(2)

    2011

  44. [52]

    Nils Reimers and Iryna Gurevych. 2019. https://doi.org/10.18653/v1/D19-1410 Sentence- BERT : Sentence embeddings using S iamese BERT -networks . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference...

    2019 doi

  45. [53]

    Stuart Rose, Dave Engel, Nick Cramer, and Wendy Cowley. 2010. Automatic keyword extraction from individual documents. Text mining: applications and theory, pages 1--20

    2010

  46. [54]

    Maarten Sap, Dallas Card, Saadia Gabriel, Yejin Choi, and Noah A. Smith. 2019. https://doi.org/10.18653/v1/P19-1163 The risk of racial bias in hate speech detection . In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 1668--1678, ...

    2019 doi

  47. [55]

    Jennifer Saul. 2018. Dogwhistles, political manipulation, and philosophy of language. New work on speech acts, 360:84

    2018

  48. [56]

    Jingbo Shang, Jialu Liu, Meng Jiang, Xiang Ren, Clare R Voss, and Jiawei Han. 2017. https://arxiv.org/abs/1702.04457 Automated phrase mining from massive text corpora . Preprint, arXiv:1702.04457

    2017 arXiv

  49. [57]

    Karen Spärck Jones. 1972. https://doi.org/10.1108/eb026526 A statistical interpretation of term specificity and its application in retrieval . Journal of Documentation, 28(1):11--21

    1972 doi

  50. [58]

    Jherez Taylor, Melvyn Peignon, and Yi-Shin Chen. 2017. https://arxiv.org/abs/1711.10093 Surfacing contextual hate speech words within social media . Preprint, arXiv:1711.10093

    2017 arXiv

  51. [59]

    i am the law and order candidate

    Brian P Tilley and 1 others. 2020. “i am the law and order candidate”: A content analysis of donald trump’s race-baiting dog whistles in the 2016 presidential campaign. Psychology, 11(12):1941

    2020

  52. [60]

    José Ramón Torices. 2021. https://www.jstor.org/stable/27073752 Understanding dogwhistles politics - comprender la política de los silbatos para perros . Theoria: An International Journal for Theory, History and Foundations of Science, 36(3):321--339

    2021

  53. [61]

    Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, Dan Bikel, Lukas Blecher, Cristian Canton Ferrer, Moya Chen, Guillem Cucurull, David Esiobu, Jude Fernandes, Jeremy Fu, W...

    2023 arXiv

  54. [62]

    Bertie Vidgen, Tristan Thrush, Zeerak Waseem, and Douwe Kiela. 2021. https://doi.org/10.18653/v1/2021.acl-long.132 Learning from the worst: Dynamically generated datasets to improve online hate detection . In Proceedings of the 59th Annual Meeting of the Association for Comput...

    2021 doi

  55. [63]

    Zeerak Waseem, Thomas Davidson, Dana Warmsley, and Ingmar Weber. 2017. https://doi.org/10.18653/v1/W17-3012 Understanding abuse: A typology of abusive language detection subtasks . In Proceedings of the First Workshop on Abusive Language Online, pages 78--84, Vancouver, BC, Ca...

    2017 doi

  56. [64]

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed Chi, Quoc Le, and Denny Zhou. 2023. https://arxiv.org/abs/2201.11903 Chain-of-thought prompting elicits reasoning in large language models . Preprint, arXiv:2201.11903

    2023 arXiv

  57. [65]

    Rachel Wetts and Robb Willer. 2019. Who is called by the dog whistle? experimental evidence that racial resentment and political ideology condition responses to racially encoded messages. Socius, 5:2378023119866268

    2019

  58. [66]

    Hao Yang, Xiulin Ma, Kun Du, Zhou Li, Haixin Duan, Xiaodong Su, Guang Liu, Zhifeng Geng, and Jianping Wu. 2017. https://doi.org/10.1109/SP.2017.11 How to learn klingon without a dictionary: Detection and measurement of black keywords used by the underground economy . In 2017 I...

    2017 doi

  59. [67]

    Kan Yuan, Haoran Lu, Xiaojing Liao, and XiaoFeng Wang. 2018. Reading thieves' cant: automatically identifying and understanding dark jargons from cybercrime marketplaces. In 27th USENIX Security Symposium (USENIX Security 18), pages 1027--1041

    2018

  60. [68]

    Min Zhang, Jianfeng He, Taoran Ji, and Chang-Tien Lu. 2024. https://arxiv.org/abs/2402.11406 Don't go to extremes: Revealing the excessive sensitivity and calibration limitations of llms in implicit hate speech detection . Preprint, arXiv:2402.11406

    2024 arXiv

  61. [69]

    Kangzhi Zhao, Yong Zhang, Chunxiao Xing, Weifeng Li, and Hsinchun Chen. 2016. https://doi.org/10.1109/ISI.2016.7745450 Chinese underground market jargon analysis based on unsupervised learning . In 2016 IEEE Conference on Intelligence and Security Informatics (ISI), pages 97--102

    2016

  62. [70]

    Wanzheng Zhu and Suma Bhat. 2021. https://doi.org/10.18653/v1/2021.findings-emnlp.16 Euphemistic phrase detection by masked language model . In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 163--168, Punta Cana, Dominican Republic. Association fo...

    2021 doi

  63. [71]

    Wanzheng Zhu, Hongyu Gong, Rohan Bansal, Zachary Weinberg, Nicolas Christin, Giulia Fanti, and Suma Bhat. 2021. https://arxiv.org/abs/2103.16808 Self-supervised euphemism detection and identification for content moderation . Preprint, arXiv:2103.16808

    2021 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.