Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Understanding Toxic Interaction Across User and Video Clusters in Social Video Platforms

Qiao Wang, Liang Liu, Mitsuo Yoshida

TL;DR

The paper addresses toxic interactions on social video platforms by modeling users and videos as a bipartite interaction matrix and applying K-means clustering to both sides after normalization and dimensionality reduction. It integrates barrage and comment modalities with textual signals (toxicity and sentiment) to identify stable video and user clusters and examine how exposure, interaction style, and content relate to toxicity. Key findings show that large-scale video exposure correlates with higher toxicity, while certain user clusters exhibit longer, more constructive messages; governance implications propose targeted moderation and engagement strategies. The work provides a framework for understanding platform governance in hybrid social–video ecosystems and highlights the importance of considering structural context in toxicity research.

Abstract

Social video platforms shape how people access information, while recommendation systems can narrow exposure and increase the risk of toxic interaction. Previous research has often examined text or users in isolation, overlooking the structural context in which such toxic interactions occur. Without considering who interacts with whom and around what content, it is difficult to explain why negative expressions cluster within particular communities. To address this issue, this study focuses on the Chinese social video platform Bilibili, incorporating video-level information as the environment for user expression, modeling users and videos in an interaction matrix. After normalization and dimensionality reduction, we perform separate clustering on both sides of the video-user interaction matrix with K-means. Cluster assignments facilitate comparisons of user behavior, including message length, posting frequency, and source (barrage and comment), as well as textual features such as sentiment and toxicity, and video attributes defined by uploaders. Such a clustering approach integrates structural ties with content signals to identify stable groups of videos and users. We find clear stratification in interaction style (message length, comment ratio) across user clusters, while sentiment and toxicity differences are weak or inconsistent across video clusters. Across video clusters, viewing volume exhibits a clear hierarchy, with higher exposure groups concentrating more toxic expressions. For such a group, platforms should require timely intervention during periods of rapid growth. Across user clusters, comment ratio and message length form distinct hierarchies, and several clusters with longer and comment-oriented messages exhibit lower toxicity. For such groups, platforms should strengthen mechanisms that sustain rational dialogue and encourage engagement across topics.

Understanding Toxic Interaction Across User and Video Clusters in Social Video Platforms

TL;DR

The paper addresses toxic interactions on social video platforms by modeling users and videos as a bipartite interaction matrix and applying K-means clustering to both sides after normalization and dimensionality reduction. It integrates barrage and comment modalities with textual signals (toxicity and sentiment) to identify stable video and user clusters and examine how exposure, interaction style, and content relate to toxicity. Key findings show that large-scale video exposure correlates with higher toxicity, while certain user clusters exhibit longer, more constructive messages; governance implications propose targeted moderation and engagement strategies. The work provides a framework for understanding platform governance in hybrid social–video ecosystems and highlights the importance of considering structural context in toxicity research.

Abstract

Social video platforms shape how people access information, while recommendation systems can narrow exposure and increase the risk of toxic interaction. Previous research has often examined text or users in isolation, overlooking the structural context in which such toxic interactions occur. Without considering who interacts with whom and around what content, it is difficult to explain why negative expressions cluster within particular communities. To address this issue, this study focuses on the Chinese social video platform Bilibili, incorporating video-level information as the environment for user expression, modeling users and videos in an interaction matrix. After normalization and dimensionality reduction, we perform separate clustering on both sides of the video-user interaction matrix with K-means. Cluster assignments facilitate comparisons of user behavior, including message length, posting frequency, and source (barrage and comment), as well as textual features such as sentiment and toxicity, and video attributes defined by uploaders. Such a clustering approach integrates structural ties with content signals to identify stable groups of videos and users. We find clear stratification in interaction style (message length, comment ratio) across user clusters, while sentiment and toxicity differences are weak or inconsistent across video clusters. Across video clusters, viewing volume exhibits a clear hierarchy, with higher exposure groups concentrating more toxic expressions. For such a group, platforms should require timely intervention during periods of rapid growth. Across user clusters, comment ratio and message length form distinct hierarchies, and several clusters with longer and comment-oriented messages exhibit lower toxicity. For such groups, platforms should strengthen mechanisms that sustain rational dialogue and encourage engagement across topics.

Paper Structure

This paper contains 15 sections, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Methodology
  4. Data Description
  5. Toxicity and Sentiment Analysis
  6. Statistical and Analytical Methods
  7. Results
  8. Overview of Clustering Results
  9. Video Clustering Results
  10. User Clustering Results
  11. Discussion
  12. Interpretation of Findings
  13. Governance Implications
  14. Limitations and Future Directions
  15. Conclusion

Figures (4)

  • Figure 1: t-SNE visualization of video clusters. Colors indicate different cluster IDs in the two-dimensional embedding space.
  • Figure 2: t-SNE visualization of user clusters. Colors indicate different cluster IDs in the two-dimensional embedding space.
  • Figure 3: Video category distribution across video clusters. Bar plots show the distribution of video categories within each video cluster. Entertainment categories dominate, while specific clusters (e.g., cluster 6) show higher proportions of Social/Law/Psych and PC/Console games, reflecting thematic differentiation.
  • Figure 4: Video category distribution across user clusters. The stacked bar chart shows the relative share of video categories, the top 10, and Others, within each user cluster. Entertainment talk dominates across clusters, but clusters 3 and 6 show higher proportions of Social/Law/Psych or Variety show content, reflecting distinct topical preferences.