ImpliHateVid: A Benchmark Dataset and Two-stage Contrastive Learning Framework for Implicit Hate Speech Detection in Videos

2025-08-07

AI Summary

Implicit hate speech detection in videos remains underexplored, hindered by the absence of large-scale, annotated multimodal benchmarks. Method: We introduce ImpliHateVid—the first large-scale, multimodal dataset specifically designed for implicit hate speech in video—and propose a two-stage contrastive learning framework. Stage one employs modality-specific encoders to extract features from audio, text, visual content, and affective/sentiment/caption modalities. Stage two performs cross-modal contrastive learning to refine joint representations, enhancing modeling of complex implicit hateful semantics. Contribution/Results: Our approach integrates multimodal deep learning, cross-encoder training, and fine-grained semantic analysis. It achieves state-of-the-art performance on both ImpliHateVid and the existing HateMM benchmark, significantly outperforming prior baselines. Empirical results validate both the utility of the proposed dataset and the generalizability and effectiveness of the framework.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

The existing research has primarily focused on text and image-based hate speech detection, video-based approaches remain underexplored. In this work, we introduce a novel dataset, ImpliHateVid, specifically curated for implicit hate speech detection in videos. ImpliHateVid consists of 2,009 videos comprising 509 implicit hate videos, 500 explicit hate videos, and 1,000 non-hate videos, making it one of the first large-scale video datasets dedicated to implicit hate detection. We also propose a novel two-stage contrastive learning framework for hate speech detection in videos. In the first stage, we train modality-specific encoders for audio, text, and image using contrastive loss by concatenating features from the three encoders. In the second stage, we train cross-encoders using contrastive learning to refine multimodal representations. Additionally, we incorporate sentiment, emotion, and caption-based features to enhance implicit hate detection. We evaluate our method on two datasets, ImpliHateVid for implicit hate speech detection and another dataset for general hate speech detection in videos, HateMM dataset, demonstrating the effectiveness of the proposed multimodal contrastive learning for hateful content detection in videos and the significance of our dataset.

Problem

• Detecting implicit hate speech in videos using multimodal data
• Addressing lack of large-scale video datasets for hate detection
• Improving hate speech detection with contrastive learning framework

Innovation

• Two-stage contrastive learning for video hate detection
• Multimodal encoders for audio, text, image
• Sentiment, emotion, caption features enhance detection