Research Area:  Machine Learning
Expression of emotion is an indicator that can contribute to the detection of mental health-related disorders. Suicide causes death to many people around the globe, and despite the suicide prevention strategies that have been employed over the years, only a few studies have explored the role of emotions in predicting suicidal behavior on social media platforms. This study explored the role of emotions from Twitter messages in detecting suicide-related content. We extracted and analyzed the characteristics of Twitter users sentiment and behavior response (anger, fear, sadness, joy, positive, and negative) using NRC Affect Intensity Lexicon and SentiStrength techniques. A semi-supervised learning method was applied using the YATSI classifier or “Yet Another Two-Stage Idea” to efficiently recognize suicide-related tweets. The results showed that tweets associated with suicide content were exclusively related to fear, sadness, and negative sentiments. The classification results showed the potential of emotions in facilitating the detection of suicide-related content online. Our findings offer valuable insights into ongoing research on the prevention of suicide risk and other mental-related disorders on Twitter. The proposed mechanism can contribute to the development of clinical decision support systems that deal with evidence-based guidelines and generate customized recommendations.
Author(s) Name:  Samer Muthana Sarsam, Hosam Al-Samarraie, Ahmed Ibrahim Alzahrani, Waleed Alnumay, Andrew Paul Smith
Journal name:  Biomedical Signal Processing and Control
Publisher name:  ELSEVIER
Volume Information:  Volume 65, March 2021, 102355
Paper Link:   https://www.sciencedirect.com/science/article/abs/pii/S1746809420304651