Towards Explainable Hate Speech Detection (ACl Anthology)

Natural language processing (NLP) activities are now much more accurate and efficient because to recent developments in deep learning. However, a significant disadvantage of these models is that they frequently call for a significant amount of processing power. Cost-effective NLP solutions can be achieved by simplifying deep learning architectures and investigating more straightforward but efficient methods. Finding out how a specific task is completed is another step toward explainable AI. The researchers decided to use the hate speech identification challenge for the current investigation. By presenting a model that uses a weighted sum of valence, arousal, and dominance (VAD) scores for classification, they tackle hate speech detection. They examine hate speech and non-hate speech words according to their individual and combined VAD-values in order to identify the best weights and categorization schemes.

https://aclanthology.org/2025.findings-acl.667