Classification of Helpful Comments on Online Suicide Watch Forums.

Area of Science:

• Computational Linguistics
• Mental Health Technology
• Social Media Analysis

Background:

• Reddit's Suicide Watch (SW) subreddit is a major online forum for mental health discussions.
• Timely moderation of user comments expressing suicidal thoughts is challenging due to volume and global accessibility.
• Current Reddit comment ranking lacks a suicide prevention (SP) helpfulness metric.

Purpose of the Study:

• To develop and evaluate an automated system for identifying helpful comments within online suicide prevention discussions.
• To assist human moderators and provide immediate feedback on comment relevance for users seeking suicide prevention support.
• To offer insights for suicide prevention researchers on online communication patterns.

Main Methods:

• Utilized a dataset of 3000 real comments from the SW subreddit.
• Incorporated supervised learning with content-based features, including n-grams, word psychometric scores, and discourse relation graphs.
• Obtained expert judgments from SP researchers on comment helpfulness within their original post contexts.

Main Results:

• Achieved high F-scores (approximately 80-90%) in classifying helpful comments.
• Demonstrated the potential of machine learning for complementing moderation in SW forums.
• Highlighted the complexity and need for further research in assessing online mental health comment helpfulness.

Conclusions:

• Machine learning models can effectively identify helpful comments in online suicide prevention contexts.
• Automated systems can support human moderators and enhance user experience on mental health forums.
• Further research is needed to refine the understanding and assessment of helpfulness in online mental health discourse.