mlsanityData to 7 Oct 2026

Challenges for Toxic Comment Classification: An In-Depth Error Analysis

arXiv 1809.07572 ↗First posted 20 Sep 2018cs.CLcited by 55

Not on the map. Why?

Abstract (arXiv)

Toxic comment classification has become an active research field with many recently proposed approaches. However, while these approaches address some of the task's challenges others still remain unsolved and directions for further research are needed. To this end, we compare different deep learning and shallow approaches on a new, large comment dataset and propose an ensemble that outperforms all individual models. Further, we validate our findings on a second dataset. The results of the ensemble enable us to perform an extensive error analysis, which reveals open challenges for state-of-the-art methods and directions towards pending future research. These challenges include missing paradigmatic context and inconsistent dataset labels.

Key facts. Cited by 55 papers in the map (data through 7 Oct 2026). First posted on arXiv 20 Sep 2018.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 13; 2022: 6; 2023: 16; 2024: 11; 2025: 7; 2026: 2.

55 papers.

Sort:

By area:

  1. 2023Trustworthy LLMs: a Survey and Guideline for Evaluating Large Language Models' Alignmentcited by 394
  2. 2022Jury Learning: Integrating Dissenting Voices into Machine Learning Modelscited by 115
  3. 2022Pile of Law: Learning Responsible Data Filtering from the Law and a 256GB Open-Source Legal Datasetcited by 113
  4. 2021Self-Attention Between Datapoints: Going Beyond Individual Input-Output Pairs in Deep Learningcited by 112
  5. 2021Towards generalisable hate speech detection: a review on obstacles and solutionscited by 66
  6. 2022Multilingual HateCheck: Functional Tests for Multilingual Hate Speech Detection Modelscited by 51
  7. 2023Multi-lingual and Multi-cultural Figurative Language Understandingcited by 47
  8. 2021Detecting Cross-Geographic Biases in Toxicity Modeling on Social Mediacited by 31
  9. 2023Unmasking and Improving Data Credibility: A Study with Datasets for Training Harmless Language Modelscited by 26
  10. 2024Model Surgery: Modulating LLM's Behavior Via Simple Parameter Editingcited by 23
  11. 2021Data Expansion using Back Translation and Paraphrasing for Hate Speech Detectioncited by 20
  12. 2024NLP Systems That Can't Tell Use from Mention Censor Counterspeech, but Teaching the Distinction Helpscited by 19
38 more
  1. 2021Measuring and Improving Model-Moderator Collaboration using Uncertainty Estimationcited by 18
  2. 2023Towards Safer Generative Language Models: A Survey on Safety Risks, Evaluations, and Improvementscited by 14
  3. 2024MultiParaDetox: Extending Text Detoxification with Parallel Data to New Languagescited by 13
  4. 2023Characterizing Large Language Model Geometry Helps Solve Toxicity Detection and Generationcited by 11
  5. 2021Deciphering Implicit Hate: Evaluating Automated Detection Algorithms for Multimodal Hatecited by 8
  6. 2022Automated Identification of Toxic Code Reviews Using ToxiCRcited by 8
  7. 2025From LLM-anation to LLM-orchestrator: Coordinating Small Models for Data Labelingcited by 8
  8. 2024Large Language Models can be Strong Self-Detoxifierscited by 7
  9. 2024Harmful YouTube Video Detection: A Taxonomy of Online Harm and MLLMs as Alternative Annotatorscited by 7
  10. 2022Hidden behind the obvious: misleading keywords and implicitly abusive language on social mediacited by 6
  11. 2024DefVerify: Do Hate Speech Models Reflect Their Dataset's Definition?cited by 6
  12. 2021Predicting Different Types of Subtle Toxicity in Unhealthy Online Conversationscited by 5
  13. 2021Limitations of machine learning for building energy prediction: ASHRAE Great Energy Predictor III Kaggle competition error analysiscited by 5
  14. 2021From Theories on Styles to their Transfer in Text: Bridging the Gap with a Hierarchical Surveycited by 5
  15. 2022Leveraging a New Spanish Corpus for Multilingual and Crosslingual Metaphor Detectioncited by 5
  16. 2023ToxBuster: In-game Chat Toxicity Buster with BERTcited by 5
  17. 2023Adding guardrails to advanced chatbotscited by 5
  18. 2023Characterizing Information Seeking Events in Health-Related Social Discoursecited by 5
  19. 2025Unified Game Moderation: Soft-Prompting and LLM-Assisted Label Transfer for Resource-Efficient Toxicity Detectioncited by 5
  20. 2023Cross-Domain Toxic Spans Detectioncited by 4
  21. 2023Towards Detecting Contextual Real-Time Toxicity for In-Game Chatcited by 4
  22. 2023A Taxonomy of Rater Disagreements: Surveying Challenges & Opportunities from the Perspective of Annotating Online Toxicitycited by 4
  23. 2024BadFair: Backdoored Fairness Attacks with Group-conditioned Triggerscited by 4
  24. 2024Unaligning Everything: Or Aligning Any Text to Any Image in Multimodal Modelscited by 3
  25. 2025Toxicity in Online Platforms and AI Systems: A Survey of Needs, Challenges, Mitigations, and Future Directionscited by 3
  26. 2021SS-BERT: Mitigating Identity Terms Bias in Toxic Comment Classification by Utilising the Notion of "Subjectivity" and "Identity Terms"cited by 2
  27. 2023Safer Conversational AI as a Source of User Delightcited by 2
  28. 2024The Constant in HATE: Analyzing Toxicity in Reddit across Topics and Languagescited by 2
  29. 2024StopHC: A Harmful Content Detection and Mitigation Architecture for Social Media Platformscited by 2
  30. 2025A Survey of Machine Learning Models and Datasets for the Multi-label Classification of Textual Hate Speech in Englishcited by 2
  31. 2025K/DA: Automated Data Generation Pipeline for Detoxifying Implicitly Offensive Language in Koreancited by 2
  32. 2021UPB at SemEval-2021 Task 5: Virtual Adversarial Training for Toxic Spans Detectioncited by 1
  33. 2023Healing Unsafe Dialogue Responses with Weak Supervision Signalscited by 1
  34. 2023Studying Socially Unacceptable Discourse Classification (SUD) through different eyes: "Are we on the same page ?"cited by 1
  35. 2024Grounding Toxicity in Real-World Events across Languagescited by 1
  36. 2021Detecting Hostile Posts using Relational Graph Convolutional Networkcited by 0
  37. 2021UoT-UWF-PartAI at SemEval-2021 Task 5: Self Attention Based Bi-GRU with Multi-Embedding Representation for Toxicity Highlightercited by 0
  38. 2023Classification of social media Toxic comments using Machine learning modelscited by 0

50 most cited of 55 All 55 (JSON)