Implicit Hate Speech Detection Based on Sentiment and Common Sense Knowledge Enhancement

Zhong Zhang, Yushan Jiang, Weiwei Ouyang, Lei Tian, Xuechen Zhao2025 · DOI

Abstract

We propose a novel method for implicit hate speech detection that integrates sentiment and common sense knowledge. By using a pre-trained language model to learn the textual content representation of the original tweets, and combining the two dimensions of sentiment and hate for contrastive learning, the multi-dimensional hate expression features are learned. Furthermore, to fully incorporate relevant external knowledge, our method retrieves common sense knowledge concepts from the semantic network to combine with the text to construct a directed isomorphic graph for each sample. Finally, an attention mechanism is used to fuse the learned feature representations from the two modules for implicit hate speech detection. Experimental results demonstrate that our method performs well in terms of both accuracy and efficiency, offering an effective solution for detecting implicit hate speech on social media.
BibTeX
@misc{zhang2025implicit,
  author = {Zhong Zhang and Yushan Jiang and Weiwei Ouyang and Lei Tian and Xuechen Zhao},
  title = {{Implicit Hate Speech Detection Based on Sentiment and Common Sense Knowledge Enhancement}},
  year = {2025},
  doi = {10.1109/iscipt67144.2025.11264958},
}

Loading related content...

Figure