Hate Speech and Sentiment of YouTube Video Comments From Public and Private Sources Covering the Israel-Palestine Conflict

TL;DR

This work addresses HS detection and sentiment analysis in YouTube comments about the Israel-Palestine conflict, focusing on German-language discourse from public and private news sources. It builds a carefully annotated dataset of $4{,}983$ comments and trains LR and SVM classifiers achieving AUROC from $0.83$ to $0.90$, subsequently applying them to field data to compare HS prevalence and sentiment across source types. The study finds higher HS in public sources ($40.4\%$) than private sources ($31.6\%$) and predominantly neutral sentiment, with public data showing stronger orientation toward Israel and Palestine. These insights inform content moderation and highlight the dynamic interplay between real-world events and online discourse, while outlining limitations and directions for longitudinal, multi-modal follow-up work.

Abstract

This study explores the prevalence of hate speech (HS) and sentiment in YouTube video comments concerning the Israel-Palestine conflict by analyzing content from both public and private news sources. The research involved annotating 4983 comments for HS and sentiments (neutral, pro-Israel, and pro-Palestine). Subsequently, machine learning (ML) models were developed, demonstrating robust predictive capabilities with area under the receiver operating characteristic (AUROC) scores ranging from 0.83 to 0.90. These models were applied to the extracted comment sections of YouTube videos from public and private sources, uncovering a higher incidence of HS in public sources (40.4%) compared to private sources (31.6%). Sentiment analysis revealed a predominantly neutral stance in both source types, with more pronounced sentiments towards Israel and Palestine observed in public sources. This investigation highlights the dynamic nature of online discourse surrounding the Israel-Palestine conflict and underscores the potential of moderating content in a politically charged environment.

Figures (4)

• Figure 1: Framework of our work.
• Figure 2: Word cloud of field data.
• Figure 3: Hate progression over time.
• Figure 4: Sentiment progression over time.