Revealing Temporal Label Noise in Multimodal Hateful Video Classification

2025-08-06

AI Summary

Online hate video detection suffers from temporal label noise due to coarse-grained, video-level annotations: videos labeled as “hateful” often contain extensive non-hateful segments, leading to semantic ambiguity and model decision bias. This work is the first to systematically characterize the detrimental impact of temporal label noise on multimodal classification performance. We propose a fine-grained, timestamp-accurate video segment cropping method and conduct controlled experiments and exploratory analyses on the HateMM and MultiHateClip datasets. Results demonstrate that label noise substantially reduces model confidence, distorts decision boundaries, and reveals the strong contextual dependency and temporal continuity inherent in hateful expression. Our findings advance temporal-aware modeling paradigms and provide both theoretical grounding and practical guidance for constructing high-quality, fine-grained benchmarks and robust hate detection models.

Technology Category

• Computer Vision: Video Understanding & Activity Analysis
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs

Abstract

The rapid proliferation of online multimedia content has intensified the spread of hate speech, presenting critical societal and regulatory challenges. While recent work has advanced multimodal hateful video detection, most approaches rely on coarse, video-level annotations that overlook the temporal granularity of hateful content. This introduces substantial label noise, as videos annotated as hateful often contain long non-hateful segments. In this paper, we investigate the impact of such label ambiguity through a fine-grained approach. Specifically, we trim hateful videos from the HateMM and MultiHateClip English datasets using annotated timestamps to isolate explicitly hateful segments. We then conduct an exploratory analysis of these trimmed segments to examine the distribution and characteristics of both hateful and non-hateful content. This analysis highlights the degree of semantic overlap and the confusion introduced by coarse, video-level annotations. Finally, controlled experiments demonstrated that time-stamp noise fundamentally alters model decision boundaries and weakens classification confidence, highlighting the inherent context dependency and temporal continuity of hate speech expression. Our findings provide new insights into the temporal dynamics of multimodal hateful videos and highlight the need for temporally aware models and benchmarks for improved robustness and interpretability. Code and data are available at https://github.com/Multimodal-Intelligence-Lab-MIL/HatefulVideoLabelNoise.

Problem

• Investigating temporal label noise in hateful video classification
• Analyzing semantic overlap from coarse video-level annotations
• Examining how timestamp noise affects model decision boundaries

Innovation

• Trimmed videos using annotated timestamps
• Conducted exploratory analysis of segments
• Demonstrated timestamp noise alters model boundaries