Pith. sign in

REVIEW 3 major objections 5 minor 64 references

Sensitive Content Classification in Social Media: A Holistic Resource and Evaluation

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A new 8,000-tweet benchmark shows fine-tuned models beat off-the-shelf content moderators by 10-15 points.

desk verdict A genuinely useful multi-label sensitive-content dataset with a serious annotation-aggregation caveat, and an abstract whose 10-15% claim is not backed by the table. read the letter →

arxiv 2411.19832 v3 pith:DHLTQPE4 submitted 2024-11-29 cs.CL

Dimosthenis Antypas , Indira Sen , Carla Perez-Almendros , Jose Camacho-Collados , Francesco Barbieri This is my paper · ORCID
classification cs.CL
keywords sensitivecontentdetectionmoderationmulti-labelclassificationfine-tuningX-Sensitivedatasetsocialmediacrowdannotationoff-the-shelfLLMevaluation
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

The paper's central claim is that the bottleneck in detecting sensitive content on social media is the absence of a unified, multi-category training resource, not model size. To close this gap it introduces X-Sensitive, a manually annotated dataset of 8,000 tweets spanning six sensitive categories: conflictual language, profanity, sexually explicit material, drug-related content, self-harm, and spam. Fine-tuning an 8-billion-parameter language model on this dataset reaches 85.6 macro-F1 in binary detection and 69.8 in multi-label detection, outperforming open and proprietary off-the-shelf models by 10-15 points overall. A 355-million-parameter model trained on social media comes within about 2 points of the best model, which the paper reads as evidence that specialised data, not scale, drives performance. If this holds, the dataset offers a practical, privacy-preserving alternative to moderation APIs that cannot be customised to specific sensitive categories.

What carries the argument

The central object is the X-Sensitive dataset itself: 8,000 tweets, multi-label, covering conflictual language, profanity, sexually explicit material, drug-related content, self-harm, and spam. It is constructed to avoid the usual weaknesses of keyword-scraped sets by expanding seed word lists with tweet-trained word embeddings, manually filtering low-precision terms, and holding out roughly half the test set so those tweets share no collection keywords with the training split. The carrying mechanism is the combination of a consistent taxonomy drawn from platform community guidelines, a recall-oriented annotation rule, and the use of the same collection and labelling pipeline across all six categories; this is what lets one model be trained and evaluated on all categories at once rather than on separate, inconsistently built datasets.

What would settle it

Re-annotate a random sample of about 200 test tweets with a panel of five trained moderators using a strict majority rule, then compare their labels to X-Sensitive. If per-category agreement is close to the original annotators' internal agreement (alpha around 0.49-0.56), the reported F1 differences between systems are partly measuring label noise, and the rankings could change on a cleaner gold standard.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that a purpose-built, multi-label dataset makes specialised sensitive-content classifiers substantially better than general-purpose moderation systems. X-Sensitive labels 8,000 tweets with six categories, collected through embedding-expanded keyword lists and annotated by crowd workers using a recall-oriented rule: a tweet receives a label when at least one annotator answers yes and the rest answer yes or not sure, so no direct veto is allowed. The paper reports that fine-tuned Llama-3-8B is the best system overall, with macro-F1 (the average of per-class F1 scores) of 85.6 in the binary setting and 69.8 in the multi-label setting; gpt-4o in a few-shot setting reaches 83.3 and 67.9, while out-of-the-box moderation APIs fall further behind. The authors also find that fine-tuning helps every model tested and that a much smaller social-media-pretrained model is only about 2 points behind, which supports their argument that the value lies in the dataset itself.

Load-bearing premise

The load-bearing premise is that the crowd labels in X-Sensitive are accurate enough to serve as ground truth, even though annotator agreement is moderate (Krippendorff's alpha 0.49 multi-label, 0.56 binary) and a single 'yes' with no outright 'no' is enough to assign a label.

Editorial extensions

If this is right

  • Fine-tuned models built on X-Sensitive can run locally, giving researchers a privacy-preserving alternative to sending data to external moderation APIs.
  • The 10-15 point gap between fine-tuned and off-the-shelf systems implies that general-purpose and commercial models need category-specific training data before they can be relied on for moderation of self-harm, drug, or spam content.
  • The close performance of the 355-million-parameter model shows that a moderately sized, social-media-pretrained model is a practical option when compute is limited.
  • The multi-label design captures overlaps such as profanity combined with conflictual or sexually explicit language, so a single system can flag multiple harms in one tweet.
  • The keyword-disjoint test split gives a more realistic estimate of performance on unseen vocabulary than earlier keyword-based datasets.

Reading between the lines

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

  • Beyond the paper: the no-veto label rule deliberately trades precision for recall, so a model trained on X-Sensitive will tend to over-flag borderline tweets, meaning more false positives for human moderators to review.
  • Beyond the paper: because annotator agreement is only moderate (alpha 0.49 multi-label, 0.56 binary), some part of the measured gap between fine-tuned and off-the-shelf models could be the fine-tuned models learning the labelling rule rather than the categories themselves; an expert re-annotation study would separate these.
  • Beyond the paper: the same collection pipeline could be ported to other languages or platforms to test whether the fine-tuning advantage persists, since the paper itself notes that the dataset is English-only and drawn from a single platform.
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. The paper introduces X-Sensitive, a manually annotated multi-label dataset of 8,000 English tweets covering six sensitive content categories (conflictual language, profanity, sexually explicit content, drug-related content, self-harm, and spam). Tweets were collected using a keyword-expansion technique and annotated by crowdsourced workers, with labels aggregated via a permissive rule requiring at least one 'yes' and no direct 'no'. The authors evaluate fine-tuned models (RoBERTa, TimeLMs, Llama-3-8B) and off-the-shelf systems (zero-/few-shot LLMs and moderation APIs) in binary and multi-label settings, reporting that fine-tuning yields substantial gains and that the best model, fine-tuned Llama-3-8B, reaches macro-F1 scores of 85.6 (binary) and 69.8 (multi-label). The paper also analyzes annotator demographics and category overlap, and releases the dataset, models, and annotator-level labels.

Significance. If the label-quality concerns are addressed, X-Sensitive would constitute a valuable unified resource for sensitive content moderation, covering under-studied categories such as self-harm and drugs alongside more commonly studied categories. The open release of the data, fine-tuned models, and annotator-level labels is a notable strength, as is the breadth of the evaluation across different model families and API-based systems. However, the central performance claim of a 10-15% advantage over proprietary LLMs is not fully supported by the reported numbers, and the low inter-annotator agreement combined with a permissive label aggregation rule raises questions about whether fine-tuned models are learning a robust notion of sensitivity or merely reproducing a noisy annotation convention. The paper's demographic analysis and class-overlap statistics are informative and add to the resource's potential utility.

major comments (3)
  1. [§3.3] The label aggregation rule assigns a positive label when at least one annotator answers 'yes' and the remaining annotators answer 'yes' or 'not sure', with no direct 'no'. This is a very permissive rule, especially given that Krippendorff's alpha is 0.49 in the multi-label setting and 0.56 in the binary setting. With three annotators, a tweet with one 'yes' and two 'not sure' becomes a positive training and test instance. Since the fine-tuned models are trained to reproduce this mapping, their advantage over off-the-shelf LLMs may partly reflect the models' adaptation to this particular label convention rather than to a more robust notion of sensitive content. The paper should report results on high-agreement subsets (e.g., unanimous or majority-vote labels) and/or evaluate with the annotator-level labels now linked in the Ethics statement, to demonstrate that the main findings are not an artifact of this permissive aggregation rule.
  2. [Abstract and Table 2] The abstract states that proprietary OpenAI models 'underperform by 10-15% overall', but the numbers in Table 2 do not consistently support this. The largest gap between the best fine-tuned model (llama3-8b) and the best zero-shot model (gpt-4o) is 9.9 points in the binary setting (85.6 vs 75.7) and 4.9 points in the multi-label setting (69.8 vs 64.9). No confidence intervals, standard deviations, or statistical significance tests are reported, and the multi-label gap is far below 10%. The claim should be revised to match the reported numbers, and ideally supported with variance estimates or significance tests.
  3. [§4.2.3 and Table 2] The comparison with out-of-the-box systems is weakened by taxonomy mismatch. Perspective and OpenAI moderation API categories are mapped to X-Sensitive categories, but the mapping is partial (e.g., OpenAI's 'violence' categories are not mapped to Conflictual, and Perspective's 'THREAT' and 'INSULT' are not mapped). The paper acknowledges that the systems are 'not fully comparable', but the headline conclusion is nonetheless based on these comparisons. Please provide a complete mapping table, consider evaluating on a reduced label set that is common to all systems, and report which labels were excluded and how partial matches were treated.
minor comments (5)
  1. [§3.1] The paragraph under 'Sexually Explicit Content' describes a classifier training procedure (using hatespeechdata.com and Perspective API) that is out of place in the taxonomy definition; move it to the data collection or model development section.
  2. [§5.1] The statement that llama3-8b achieves the best results 'notably achieved without any hyper-parameter tuning' is unclear given §4.2.1 says Ray Tune was used for the other fine-tuned models; clarify whether llama3-8b used default hyperparameters or a separate procedure.
  3. [Table 2] There is a typo in the column header: 'Profaninty' should be 'Profanity'. Also, consider using consistent model naming across the table and text (e.g., 'chat-gpt' vs 'gpt-4o').
  4. [§8] The Limitations section says 'we plan to release all the collected annotations', but the Ethics statement already links annotator-level labels; update the limitation to reflect that the release has occurred.
  5. [Appendix C.3] In the multi-label prompt template, '6 sexually explicit' is missing a comma; this is a minor formatting issue but should be corrected for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

Empirical benchmark with standard held-out evaluation; no circular derivation.

full rationale

The paper's central claim is that fine-tuning language models on the newly constructed X-Sensitive dataset improves detection performance relative to off-the-shelf systems. This is an empirical evaluation claim, not a derivation. The comparison systems are external (LLaMA, GPT-3.5, GPT-4o, Perspective API, OpenAI moderation API, Llama Guard), and the fine-tuned models are trained on a 6,000-tweet train split and evaluated on a disjoint 2,000-tweet test split, with class distribution stratification and a keyword-disjoint subset of the test set to probe generalization. No parameter is fitted to the test labels, and the test labels come from human annotation, not from the models being compared. The annotation aggregation rule (at least one 'yes' with no direct 'no') and the reported Krippendorff alpha values (0.49 multi-label, 0.56 binary) are data-quality and label-reliability concerns; they affect the validity of the benchmark as ground truth, but they do not make the model ranking circular, because the same labels are not used to construct the comparison systems' outputs. The paper itself acknowledges the aggregation methodology 'may be subject to differing opinions' in Section 8, which supports treating this as a limitation rather than a hidden equivalence. There are minor self-references: the TimeLMs model used in fine-tuning and the NASARI lexical specificity tool used as a sanity check are prior work by overlapping authors, but neither is load-bearing for the central claim; TimeLMs is simply one of several evaluated models, and NASARI is a descriptive statistic, not a premise of the performance comparison. No equation or construction step reduces the headline result to its own inputs. The '10-15%' underperformance statement is a loose summary of Table 2 differences rather than a formally derived quantity, and any mismatch with the table is a reporting/correctness issue, not circularity. The paper is therefore self-contained as an empirical benchmark, and no circular step meeting the required evidentiary standard is present.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim depends on dataset construction choices (keyword expansion, annotation rule) rather than mathematical derivations. The main load-bearing assumptions are the validity of platform-guideline-based taxonomy, crowd label quality, and the non-majority label aggregation rule. No new theoretical entities are introduced.

free parameters (3)
  • Keyword expansion cluster count k = not reported
    Algorithm 1 uses k clusters for keyword expansion; k is chosen by hand and affects which tweets are collected, but its value is not reported in the paper.
  • Keyword expansion neighbor counts n1 and n2 = not reported
    Algorithm 1 selects n1 closest words to each cluster mean and n2 closest neighbors per new word; both are hand-set hyperparameters that shape the retrieval vocabulary.
  • Fine-tuning hyperparameters (learning rate, epochs, warm-up) = not reported
    RoBERTa and tlm hyperparameters are optimized with Ray Tune but exact values are not listed; llama3-8b is fine-tuned without tuning. These affect reported F1 scores but are standard training choices.
assumptions (4)
  • domain assumption Platform community guidelines are a valid basis for defining sensitive content categories.
    Section 3.1 grounds the taxonomy in community guidelines of social media platforms, assuming these guidelines correspond to what moderation systems should detect.
  • domain assumption Crowd annotations from Prolific workers are reliable enough despite low inter-annotator agreement.
    Section 3.3 uses Krippendorff alpha 0.49 (multi-label) and 0.56 (binary) to validate labels; the paper treats these as acceptable based on prior studies, but this is an assumption about label quality.
  • ad hoc to paper The one-yes-plus-no-opposition label rule yields a useful ground truth rather than excessive noise.
    Section 3.3 states a label is assigned if at least one annotator says yes and the rest say yes or not sure; this non-majority rule intentionally weights recall but has no independent validation.
  • domain assumption GloVe word embeddings trained on tweets capture the vocabulary needed for keyword expansion.
    Section 3.2 relies on Pennington et al. (2014) embeddings for expanding seed keywords; if embeddings miss category-specific vocabulary, collection coverage suffers.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Sensitive Content Classification in Social Media: A Holistic Resource and Evaluation." pith.science (2026). https://pith.science/paper/DHLTQPE4

@misc{pith2026241119832,
  author       = {Pith},
  title        = {Pith review of: Sensitive Content Classification in Social Media: A Holistic Resource and Evaluation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DHLTQPE4}},
  note         = {Machine review of arXiv:2411.19832}
}
read the original abstract

The detection of sensitive content in large datasets is crucial for ensuring that shared and analysed data is free from harmful material. However, current moderation tools, such as external APIs, suffer from limitations in customisation, accuracy across diverse sensitive categories, and privacy concerns. Additionally, existing datasets and open-source models focus predominantly on toxic language, leaving gaps in detecting other sensitive categories such as substance abuse or self-harm. In this paper, we put forward a unified dataset tailored for social media content moderation across six sensitive categories: conflictual language, profanity, sexually explicit material, drug-related content, self-harm, and spam. By collecting and annotating data with consistent retrieval strategies and guidelines, we address the shortcomings of previous focalised research. Our analysis demonstrates that fine-tuning large language models (LLMs) on this novel dataset yields significant improvements in detection performance compared to open off-the-shelf models such as LLaMA, and even proprietary OpenAI models, which underperform by 10-15% overall. This limitation is even more pronounced on popular moderation APIs, which cannot be easily tailored to specific sensitive content categories, among others.

Figures

Figures reproduced from arXiv: 2411.19832 by the authors.

Figure 1
Figure 1. Overlap of classes. 1. Particularly there is a high overlap between pro￾fanity tweets and those labelled as sexual explicit content, and conflictual. In general differences between the classes are revealed even when looking at basic statistics such as the average length of tweets and the presence of emojis in them. As seen in [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Precision-Recall curve for the fine-tuned [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Guidelines provided to annotators. Tables 7 and 8 show the percentage of tweets labelled as each class by age and gender groups of coders, respectively. C Models C.1 Resources In total we estimate 112 hours used for the training of RoBERTa, tlm and llama3-8b models using a NVIDIA GeForce RTX 4090 GPU and 90 hours for inferences with the llama3-8b and llama3-70b models using an NVIDIA Quadro RTX 8000 GPU [PITH_FULL_… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Precision-Recall curve for the fine-tuned [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

64 extracted references · 36 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]

    AI@Meta. 2024. https://github.com/meta-llama/llama3/blob/main/MODEL_CARD.md Llama 3 model card

    2024

  4. [4]

    Arnav Arora, Preslav Nakov, Momchil Hardalov, Sheikh Muhammad Sarwar, Vibha Nayak, Yoan Dinkov, Dimitrina Zlatkova, Kyle Dent, Ameya Bhatawdekar, Guillaume Bouchard, et al. 2023. Detecting harmful content on online platforms: what platforms need vs. where research efforts go. ACM Computing Surveys, 56(3):1--17

    work page 2023

  5. [5]

    Dennis Assenmacher, Indira Sen, Leon Fr \"o hling, and Claudia Wagner. 2020. The end of the rehydration era the problem of sharing harmful twitter research data

    work page 2020

  6. [6]

    Gonzalo Molpeceres Barrientos, Roc \' o Alaiz-Rodr \' guez, V \' ctor Gonz \'a lez-Castro, and Andrew C Parnell. 2020. Machine learning techniques for the detection of inappropriate erotic content in text. International Journal of Computational Intelligence Systems, 13(1):591--603

    work page 2020

  7. [7]

    Priyam Basu, Tiasa Singha Roy, Soham Tiwari, and Saksham Mehta. 2021. Cyberpolice: Classification of cyber sexual harassment. In Progress in Artificial Intelligence: 20th EPIA Conference on Artificial Intelligence, EPIA 2021, Virtual Event, September 7--9, 2021, Proceedings 20, pages 701--714. Springer

    work page 2021

  8. [8]

    George Beknazar-Yuzbashev, Rafael Jim \'e nez Dur \'a n, Jesse McCrosky, and Mateusz Stalinski. 2022. Toxic content and user engagement on social media: Evidence from a field experiment. Available at SSRN 4307346

    2022

Show all 64 references
  1. [9]

    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

  2. [10]

    Cody Buntain and Jennifer Golbeck. 2015. This is your twitter on drugs: Any questions? In Proceedings of the 24th international conference on World Wide Web, pages 777--782

    2015

  3. [11]

    Jos \'e Camacho-Collados, Mohammad Taher Pilehvar, and Roberto Navigli. 2016. Nasari: Integrating explicit knowledge and corpus statistics for a multilingual representation of concepts and entities. Artificial Intelligence, 240:36--64

    2016

  4. [12]

    Robyn Caplan. 2018. Content or context moderation?

    2018

  5. [13]

    Stevie Chancellor, Michael L Birnbaum, Eric D Caine, Vincent MB Silenzio, and Munmun De Choudhury. 2019. A taxonomy of ethical tensions in inferring mental health states from social media. In Proceedings of the conference on fairness, accountability, and transparency, pages 79--88

    2019

  6. [14]

    this post will just get taken down

    Stevie Chancellor, Zhiyuan Lin, and Munmun De Choudhury. 2016 a . " this post will just get taken down" characterizing removed pro-eating disorder social media content. In Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems, pages 1157--1162

    2016

  7. [15]

    Stevie Chancellor, Zhiyuan Lin, Erica L Goodman, Stephanie Zerwas, and Munmun De Choudhury. 2016 b . Quantifying and predicting mental illness severity in online pro-eating disorder communities. In Proceedings of the 19th ACM conference on computer-supported cooperative work &...

    2016

  8. [16]

    Stevie Chancellor, Tanushree Mitra, and Munmun De Choudhury. 2016 c . Recovery amid pro-anorexia: Analysis of recovery in social media. In Proceedings of the 2016 CHI conference on human factors in computing systems, pages 2111--2123

    2016

  9. [17]

    Stevie Chancellor, Jessica Annette Pater, Trustin Clear, Eric Gilbert, and Munmun De Choudhury. 2016 d . \# thyghgapp: Instagram content moderation and lexical variation in pro-eating disorder communities. In Proceedings of the 19th ACM conference on computer-supported coopera...

    2016

  10. [18]

    Stevie Chancellor, Steven A Sumner, Corinne David-Ferdon, Tahirah Ahmad, and Munmun De Choudhury. 2021. Suicide risk and protective factors in online support forum posts: annotation scheme development and validation study. JMIR mental health, 8(11):e24471

    2021

  11. [19]

    Arijit Ghosh Chowdhury, Ramit Sawhney, Puneet Mathur, Debanjan Mahata, and Rajiv Ratn Shah. 2019. Speak up, fight back! detection of social media disclosures of sexual harassment. In Proceedings of the 2019 conference of the North American chapter of the Association for Comput...

    2019

  12. [20]

    Glen Coppersmith, Ryan Leary, Patrick Crutchley, and Alex Fine. 2018. Natural language processing of social media as screening for suicide risk. Biomedical informatics insights, 10:1178222618792860

    2018

  13. [21]

    Tao Ding, Arpita Roy, Zhiyuan Chen, Qian Zhu, and Shimei Pan. 2016. Analyzing and retrieving illicit drug-related posts from social media. In 2016 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pages 1555--1560. IEEE

    2016

  14. [22]

    Jenny Fan and Amy X Zhang. 2020. Digital juries: A civics-oriented approach to platform governance. In Proceedings of the 2020 CHI conference on human factors in computing systems, pages 1--14

    2020

  15. [23]

    participant

    Casey Fiesler and Nicholas Proferes. 2018. “participant” perceptions of twitter research ethics. Social Media+ Society, 4(1):2056305118763366

    2018

  16. [24]

    Eve Fleisig, Su Lin Blodgett, Dan Klein, and Zeerak Talat. 2024. The perspectivist paradigm shift: Assumptions and challenges of capturing human labels. arXiv preprint arXiv:2405.05860

    2024 arXiv

  17. [25]

    Paula Fortuna and S \'e rgio Nunes. 2018. A survey on automatic detection of hate speech in text. ACM Computing Surveys (CSUR), 51(4):1--30

    2018

  18. [26]

    Antigoni Founta, Constantinos Djouvas, Despoina Chatzakou, Ilias Leontiadis, Jeremy Blackburn, Gianluca Stringhini, Athena Vakali, Michael Sirivianos, and Nicolas Kourtellis. 2018. Large scale crowdsourcing and characterization of twitter abusive behavior. In Proceedings of th...

    2018

  19. [27]

    Tarleton Gillespie. 2020. Content moderation, ai, and the question of scale. Big Data & Society, 7(2):2053951720943234

    2020

  20. [28]

    Google. 2023. Perspective api. Retrieved from https://www.perspectiveapi.com/

    2023

  21. [29]

    Mitchell L Gordon, Michelle S Lam, Joon Sung Park, Kayur Patel, Jeff Hancock, Tatsunori Hashimoto, and Michael S Bernstein. 2022. Jury learning: Integrating dissenting voices into machine learning models. In Proceedings of the 2022 CHI Conference on Human Factors in Computing ...

    2022

  22. [30]

    Robert Gorwa, Reuben Binns, and Christian Katzenbach. 2020. Algorithmic content moderation: Technical and political challenges in the automation of platform governance. Big Data & Society, 7(1):2053951719897945

    2020

  23. [31]

    Maanak Gupta, CharanKumar Akiri, Kshitiz Aryal, Eli Parker, and Lopamudra Praharaj. 2023. From chatgpt to threatgpt: Impact of generative ai in cybersecurity and privacy. IEEE Access

    2023

  24. [32]

    Naveed Hussain, Hamid Turab Mirza, Ghulam Rasool, Ibrar Hussain, and Mohammad Kaleem. 2019. Spam review detection techniques: A systematic literature review. Applied Sciences, 9(5):987

    2019

  25. [33]

    Hakan Inan, Kartikeya Upasani, Jianfeng Chi, Rashi Rungta, Krithika Iyer, Yuning Mao, Michael Tontchev, Qing Hu, Brian Fuller, Davide Testuggine, et al. 2023. Llama guard: Llm-based input-output safeguard for human-ai conversations. arXiv preprint arXiv:2312.06674

    2023 arXiv

  26. [34]

    Jialun'Aaron' Jiang, Skyler Middler, Jed R Brubaker, and Casey Fiesler. 2020. Characterizing community guidelines on social media platforms. In Companion Publication of the 2020 Conference on Computer Supported Cooperative Work and Social Computing, pages 287--291

    2020

  27. [35]

    Klaus Krippendorff. 2011. Computing krippendorff's alpha-reliability

    2011

  28. [36]

    PM Lavanya and E Sasikala. 2022. Auto capture on drug text detection in social media through nlp from the heterogeneous data. Measurement: Sensors, 24:100550

    2022

  29. [37]

    Richard Liaw, Eric Liang, Robert Nishihara, Philipp Moritz, Joseph E Gonzalez, and Ion Stoica. 2018. Tune: A research platform for distributed model selection and training. arXiv preprint arXiv:1807.05118

    2018 arXiv

  30. [38]

    Luiz Henrique Quevedo Lima, Adriana Silvina Pagano, and Ana Paula Couto da Silva. 2024. Toxic content detection in online social networks: a new dataset from brazilian reddit communities. In Proceedings of the 16th International Conference on Computational Processing of Portug...

    2024

  31. [39]

    Xiao Liu, Kaixuan Ji, Yicheng Fu, Weng Lam Tam, Zhengxiao Du, Zhilin Yang, and Jie Tang. 2021. P-tuning v2: Prompt tuning can be comparable to fine-tuning universally across scales and tasks. arXiv preprint arXiv:2110.07602

    2021 arXiv

  32. [40]

    Yinhan Liu. 2019. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692, 364

    2019 arXiv

  33. [41]

    Daniel Loureiro, Francesco Barbieri, Leonardo Neves, Luis Espinosa Anke, and Jose Camacho-collados. 2022. https://doi.org/10.18653/v1/2022.acl-demo.25 T ime LM s: Diachronic language models from T witter . In Proceedings of the 60th Annual Meeting of the Association for Comput...

    2022 doi

  34. [42]

    Sourab Mangrulkar, Sylvain Gugger, Lysandre Debut, Younes Belkada, Sayak Paul, and Benjamin Bossan. 2022. Peft: State-of-the-art parameter-efficient fine-tuning methods. https://github.com/huggingface/peft

    2022

  35. [43]

    Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781

    2013 arXiv

  36. [44]

    Meena Devii Muralikumar, Yun Shan Yang, and David W McDonald. 2023. A human-centered evaluation of a toxicity detection api: Testing transferability and unpacking latent attributes. ACM Transactions on Social Computing, 6(1-2):1--38

    2023

  37. [45]

    Ayodeji Oseni, Nour Moustafa, Helge Janicke, Peng Liu, Zahir Tari, and Athanasios Vasilakos. 2021. Security and privacy for artificial intelligence: Opportunities and challenges. arXiv preprint arXiv:2102.04661

    2021 arXiv

  38. [46]

    Nedjma Ousidhoum, Xinran Zhao, Tianqing Fang, Yangqiu Song, and Dit-Yan Yeung. 2021. Probing toxic content in large pre-trained language models. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conferen...

    2021

  39. [47]

    Jeffrey Pennington, Richard Socher, and Christopher D Manning. 2014. Glove: Global vectors for word representation. In Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP), pages 1532--1543

    2014

  40. [48]

    Fabio Poletto, Valerio Basile, Manuela Sanguinetti, Cristina Bosco, and Viviana Patti. 2021. Resources and benchmark corpora for hate speech detection: a systematic review. Language Resources and Evaluation, 55:477--523

    2021

  41. [49]

    Vinodkumar Prabhakaran, Aida Mostafazadeh Davani, and Mark Diaz. 2021. On releasing annotator-level labels and information in datasets. arXiv preprint arXiv:2110.05699

    2021 arXiv

  42. [50]

    Sarah T Roberts. 2019. Behind the screen. Yale University Press

    2019

  43. [51]

    Kyle Robinson and Vijay Mago. 2022. Birds of prey: identifying lexical irregularities in spam on twitter. Wireless Networks, 28(3):1189--1196

    2022

  44. [52]

    Ramit Sawhney, Prachi Manchanda, Puneet Mathur, Rajiv Shah, and Raj Singh. 2018. Exploring and learning suicidal ideation connotations on social media with deep learning. In Proceedings of the 9th workshop on computational approaches to subjectivity, sentiment and social media...

    2018

  45. [53]

    Morgan Klaus Scheuerman, Jialun Aaron Jiang, Casey Fiesler, and Jed R Brubaker. 2021. A framework of severity for harmful content online. CSCW

    2021

  46. [54]

    Sean S Simpson, Nikki Adams, Claudia M Brugman, and Thomas J Conners. 2018. Detecting novel and emerging drug terms using natural language processing: a social media corpus study. JMIR public health and surveillance, 4(1):e7726

    2018

  47. [55]

    Michael Mesfin Tadesse, Hongfei Lin, Bo Xu, and Liang Yang. 2019. Detection of suicide ideation in social media forums using deep learning. Algorithms, 13(1):7

    2019

  48. [56]

    Vankayala Tejaswini, Korra Sathya Babu, and Bibhudatta Sahoo. 2024. Depression detection from social media text analysis using natural language processing techniques and hybrid deep learning model. ACM Transactions on Asian and Low-Resource Language Information Processing, 23(1):1--20

    2024

  49. [57]

    Samantha Thomson, Emily Kluftinger, and Jocelyn Wentland. 2018. Are you fluent in sexual emoji?: Exploring the use of emoji in romantic and sexual contexts. The Canadian Journal of Human Sexuality, 27(3):226--234

    2018

  50. [58]

    Ana-Sabina Uban and Paolo Rosso. 2020. Deep learning architectures and strategies for early detection of self-harm and depression level prediction. In CEUR workshop proceedings, volume 2696, pages 1--12. Sun SITE Central Europe

    2020

  51. [59]

    Sahana Udupa, Antonis Maronikolakis, and Axel Wisiorek. 2023. Ethical scaling for content moderation: Extreme speech and the (in) significance of artificial intelligence. Big Data & Society, 10(1):20539517231172424

    2023

  52. [60]

    Qamar Un Nisa and Rafi Muhammad. 2021. Towards transfer learning using bert for early detection of self-harm of social media users. Proceedings of the Working Notes of CLEF, pages 21--4

    2021

  53. [61]

    Bertie Vidgen, Adarsh Agrawal, Ahmed M Ahmed, Victor Akinwande, Namir Al-Nuaimi, Najla Alfaraj, Elie Alhajjar, Lora Aroyo, Trupti Bavalatti, Borhane Blili-Hamelin, et al. 2024. Introducing v0. 5 of the ai safety benchmark from mlcommons. arXiv preprint arXiv:2404.12241

    2024 arXiv

  54. [62]

    Bertie Vidgen and Leon Derczynski. 2020. Directions in abusive language training data, a systematic review: Garbage in, garbage out. Plos one, 15(12):e0243300

    2020

  55. [63]

    Michael Wiegand, Josef Ruppenhofer, and Thomas Kleinbauer. 2019. Detection of abusive language: the problem of biased datasets. In Proceedings of the 2019 conference of the North American Chapter of the Association for Computational Linguistics: human language technologies, vo...

    2019

  56. [64]

    Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, Remi Louf, Morgan Funtowicz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven Le Scao, Sylvain Gugger, Mari...

    2020 doi

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.