Sarcasm harms Turkish hate speech detection accuracy in AI models
SarcasTraps: Effect of Sarcasm on Turkish Hate Speech Detection
Computation and LanguageArtificial IntelligenceMachine Learning
Summary
Detecting hate speech is important but tricky when people use sarcasm—saying one thing but really meaning another. The authors created SarcasTraps, a special set of Turkish sentences labeled for both hate speech and sarcasm. They found that popular AI models trained to spot hate speech often miss hateful sarcasm or misclassify it. Even large language models without training struggle, though in different ways. This shows sarcasm is a real challenge for automated hate speech detection, especially in Turkish.
What this means in practice
- •For content moderation teams: Improve detection systems for Turkish hate speech by accounting for sarcastic expressions that evade current AI models.
- •For social media platform developers: Use SarcasTraps to evaluate and refine hate speech filters to better handle indirect, sarcastic harmful content in Turkish.
Authors
Niyazi Ahmet Metin, Dilara Keküllüoğlu
Abstract
Standard hate speech detection datasets often emphasize explicit harmful language, but directed harm can also be expressed indirectly, without clearly hateful wording. Sarcasm is one such form of indirect expression, as it can convey harmful intent through mock praise, exaggeration, humor, or apparent politeness. In this paper, we introduce SarcasTraps, a targeted external test set for analyzing the effect of sarcasm on Turkish hate speech detection. SarcasTraps contains 479 entries annotated along two dimensions: hate speech and sarcasm. We train classifiers using BERTurk and XLM-RoBERTa on two existing Turkish hate speech datasets and evaluate them on SarcasTraps, using it strictly as an external test set. We also evaluate GPT-4.1 and Llama-3.3-70B in a zero-shot setting on SarcasTraps. Results show that sarcasm substantially affects hate speech detection. Fine-tuned encoder models lose recall on sarcastic hate, while zero-shot LLMs maintain higher recall but suffer substantial precision drops on sarcastic examples. These findings show that both fine-tuned hate speech detectors and zero-shot LLMs struggle when sarcasm is involved, and SarcasTraps provides quantitative evidence of this challenge through a targeted Turkish test set. Due to the presence of sensitive and hateful language, SarcasTraps will be shared upon request for academic research purposes. Warning: This paper has examples from the dataset which contains offensive and hateful content towards groups of people.