Challenge: Existing toxic speech datasets lack demographic context and age data are limited . funk and its subsidiary accounts target users aged 14-29 .
Approach: a german project introduces a large-scale toxic speech dataset annotated for toxicity . the dataset includes 3,024 human-annotated and 30,024 LLM-annnotated comments . researchers used human expertise and state-of-the-art language models to label comments based on toxic keywords .
Outcome: The study combines human expertise with state-of-the-art language models to identify toxic speech categories.

Similar Papers

Toxic Language Detection in Social Media for Brazilian Portuguese: New Dataset and Multilingual Analysis (2020.aacl-main)

Copied to clipboard

Challenge: Hate speech and toxic comments are a common concern of social media platform users . identifying toxic comments is important for studying and preventing the proliferation of toxicity in social media.
Approach: They propose to use Brazilian Portuguese to analyze toxic or non-toxic tweets . they propose to analyze tweets as toxic or in different types of toxicity .
Outcome: The proposed model achieves 76% macro-F1 score using monolingual data in the binary case.
A Dataset of Offensive German Language Tweets Annotated for Speech Acts (2022.lrec-1)

Copied to clipboard

Challenge: Using speech act analysis, we analysed 600 offensive and non-offensive tweets in germany . a large body of research exists on the pragmatic characteristics of offensive language .
Approach: They analyze German offensive and non-offensive tweets and use a subset of the 2019 GermEval Shared Task on the Identification of Offensive Language dataset.
Outcome: The proposed dataset includes 600 offensive and non-offensive tweets annotated for speech acts in germany.
On the Role of Speech Data in Reducing Toxicity Detection Bias (2025.naacl-long)

Copied to clipboard

Challenge: Text toxicity detection systems produce disproportionate rates of false positives on demographic groups . toxicity classification systems often misinterpret benign group mentions as toxic .
Approach: They use group annotations to compare text-based and speech-based toxicity detection systems.
Outcome: The results show that access to speech data supports reduced bias against group mentions . the authors recommend improving classifiers, rather than transcription pipelines if possible .
Toxic, Hateful, Offensive or Abusive? What Are We Really Classifying? An Empirical Analysis of Hate Speech Datasets (2020.lrec-1)

Copied to clipboard

Challenge: a recent study shows that many definitions are being used for equivalent concepts, making most datasets incompatible.
Approach: They analyze six publicly available datasets to determine their similarity and compatibility . they propose to use Fast Text word vectors to analyze similarity between different datasets .
Outcome: The proposed model performs better on similar datasets and worse on more non-offensive samples.
A Multi-Labeled Dataset for Indonesian Discourse: Examining Toxicity, Polarization, and Demographics Information (2025.findings-acl)

Copied to clipboard

Challenge: Prior research has focused on toxicity and polarization as separate problems . extreme polarizing deepens divisions, often leading to hostility and fragmentation .
Approach: They propose to use a multi-label Indonesian dataset annotated for toxicity, polarization, and annotator demographic information to study polarizing language and toxicity.
Outcome: The proposed dataset shows that polarization cues improve toxicity classification and vice versa.
The ComMA Dataset V0.2: Annotating Aggression and Bias in Multilingual Social Media Discourse (2022.lrec-1)

Copied to clipboard

Challenge: 59,152 comments are annotated with a hierarchical, fine-grained taget marking aggression and bias of various kinds on social media platforms.
Approach: They propose to annotate a multilingual dataset with a hierarchical, fine-grained tagset marking different types of aggression and the "context" in which they occur.
Outcome: The proposed dataset contains 59,152 comments in four languages, mostly code-mixed with English.
ToxiGen: A Large-Scale Machine-Generated Dataset for Adversarial and Implicit Hate Speech Detection (2022.acl-long)

Copied to clipboard

Challenge: Toxic language detection systems often falsely flag text that contains minority group mentions as toxic . this over-reliance on spurious correlations also causes systems to struggle with detecting implicitly toxic language.
Approach: They develop a machine-generated dataset of toxic and benign statements about 13 minority groups that generates subtly toxic and harmless text with a massive pretrained language model.
Outcome: The proposed method can detect toxic and benign statements on a large scale . it can also detect hate speech on 94.5% of the toxic examples .
ModelCitizens: Representing Community Voices in Online Safety (2025.emnlp-main)

Copied to clipboard

Challenge: Existing toxic language detection models are trained on annotations that collapse diverse perspectives into a single ground truth.
Approach: They propose to augment social media posts with conversational scenarios to reflect the impact of conversational context on toxicity.
Outcome: The proposed model outperforms existing models on social media with conversational scenarios.
Exploring the Emotional Dimension of French Online Toxic Content (2024.lrec-main)

Copied to clipboard

Challenge: Emotion annotations can be used to analyze content and can be applied to content analysis.
Approach: They propose to use a corpus annotation scheme to annotate three online data sets composed of extremist, sexist and hateful messages respectively.
Outcome: The proposed method can provide new insights for content analysis and stronger empirical background for automatic content detection.
Beyond Dataset Creation: Critical View of Annotation Variation and Bias Probing of a Dataset for Online Radical Content Detection (2025.coling-main)

Copied to clipboard

Challenge: Existing datasets and models fail to address the complexities of multilingual data, authors say . detection of radical content on online platforms has become an increasingly pressing concern .
Approach: They propose a publicly available multilingual dataset annotated with radicalization levels, calls for action, and named entities in English, French, and Arabic.
Outcome: The proposed dataset is annotated with radicalization levels, calls for action, and named entities in English, French, and Arabic.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations