From Posts to Insights: A Comprehensive Analysis of the Emotion Dimensions, Engagement Level and Behavioural Patterns in Suicide-Related Tweets

Suicide is a significant global concern, causing over 700,000 deaths annually. Social media platforms are considered effective avenues for providing proactive interventions. This study comprehensively analyses 120,000 suicide-related posts on X. First, we develop SuiBERT, a fine-tuned BERT model for detecting suicidal content. We then conduct statistical analyses to examine the relationships between dimensional emotions (valence and arousal), engagement levels, and suicidality. We further employ pattern analysis to explore the account and post patterns between suicidal and non-suicidal groups. The results show that suicidal posts demonstrate lower levels of valence and arousal, along with reduced engagement levels. Valence and arousal positively correlated with engagement size but negatively with conversion rates. The findings enrich empirical insights into suicide theories and reveal the relationship between dimensional emotions and engagement levels. Practically, the pattern analysis offers valuable guidance for scholars and practitioners in developing effective suicide detection and prevention systems.