# Compositional Generalisation for Explainable Hate Speech Detection

arXiv: https://arxiv.org/abs/2506.03916 · HTML: https://mlsanity.com/paper/2506.03916/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 18 of its 36 references that are in the map.
- First posted on arXiv 4 Jun 2025.

## Builds on (most cited first, 18 of 18)

- [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://mlsanity.com/paper/1908.10084/) (2019)
- [Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detection](https://mlsanity.com/paper/2012.15761/) (2020)
- [MultiVerS: Improving scientific claim verification with weak supervision and full-document context](https://mlsanity.com/paper/2112.01640/) (2021)
- [State-of-the-art generalisation research in NLP: A taxonomy and review](https://mlsanity.com/paper/2210.03050/) (2022)
- [A BERT-Based Transfer Learning Approach for Hate Speech Detection in Online Social Media](https://mlsanity.com/paper/1910.12574/) (2019)
- [Contextualizing Hate Speech Classifiers with Post-hoc Explanation](https://mlsanity.com/paper/2005.02439/) (2020)
- [Conversational Semantic Parsing](https://mlsanity.com/paper/2009.13655/) (2020)
- [Compositional Generalization via Semantic Tagging](https://mlsanity.com/paper/2010.11818/) (2020)
- [Uncontrolled Lexical Exposure Leads to Overestimation of Compositional Generalization in Pretrained Models](https://mlsanity.com/paper/2212.10769/) (2022)
- [How Hate Speech Varies by Target Identity: A Computational Analysis](https://mlsanity.com/paper/2210.10839/) (2022)
- [HateDay: Insights from a Global Hate Speech Dataset Representative of a Day on Twitter](https://mlsanity.com/paper/2411.15462/) (2024)
- [Distinguishing rule- and exemplar-based generalization in learning systems](https://mlsanity.com/paper/2110.04328/) (2021)
- [Explainable Abuse Detection as Intent Classification and Slot Filling](https://mlsanity.com/paper/2210.02659/) (2022)
- [Improving Counterfactual Generation for Fair Hate Speech Detection](https://mlsanity.com/paper/2108.01721/) (2021)
- [Cross-Policy Compliance Detection via Question Answering](https://mlsanity.com/paper/2109.03731/) (2021)
- [Explainability and Hate Speech: Structured Explanations Make Social Media Moderators Faster](https://mlsanity.com/paper/2406.04106/) (2024)
- [Unsupervised Domain Adaptation for Hate Speech Detection Using a Data Augmentation Approach](https://mlsanity.com/paper/2107.12866/) (2021)
- [Compositional Generalization without Trees using Multiset Tagging and Latent Permutations](https://mlsanity.com/paper/2305.16954/) (2023)

## Built on by (most cited first, 1 of 1)

- [ParsHate: A Benchmark Dataset for Hate and Target Detection in Persian](https://mlsanity.com/paper/2609.16393/) (2026)

## Abstract (arXiv)

Hate speech detection is key to online content moderation, but current models struggle to generalise beyond their training data. This has been linked to dataset biases and the use of sentence-level labels, which fail to teach models the underlying structure of hate speech. In this work, we show that even when models are trained with more fine-grained, span-level annotations (e.g., "artists" is labeled as target and "are parasites" as dehumanising comparison), they struggle to disentangle the meaning of these labels from the surrounding context. As a result, combinations of expressions that deviate from those seen during training remain particularly difficult for models to detect. We investigate whether training on a dataset where expressions occur with equal frequency across all contexts can improve generalisation. To this end, we create U-PLEAD, a dataset of ~364,000 synthetic posts, along with a novel compositional generalisation benchmark of ~8,000 manually validated posts. Training on a combination of U-PLEAD and real data improves compositional generalisation while achieving state-of-the-art performance on the human-sourced PLEAD.