Suicide has emerged as a critical social health concern in contemporary society, with suicidal ideation representing individuals’ thoughts about self-harm. Factors like prolonged exposure to negative emotions or life events can contribute to these thoughts and suicide attempts. Among various suicide prevention strategies, early detection of suicidal ideation proves highly effective. The prevalence of online communication and social networking platforms offers an avenue for individuals to express their struggles and emotions, presenting an opportunity for the early detection of suicidal ideation. This project focuses on investigating online social content for the early identification of suicidal ideation.

User-generated content, particularly text posted by users, contains valuable information about individuals’ well-being and mental states. The project begins with a comprehensive content analysis to extract knowledge from suicide-related text, employing benchmarking for binary classification of suicidal ideation. This includes using feature extraction-based classifiers and deep neural networks. Recognizing the complexity of suicide motives and the variability of suicidal factors among individuals, the project integrates sentimental clues and topics from user posts. It proposes understanding the relations between these factors and posts through attention relation networks for nuanced suicidal ideation detection. The project also delves into suicidal ideation detection within private chatting scenarios. Addressing the challenge of isolated data in private chat rooms, a knowledge transfer framework is developed to train a global model for knowledge sharing among distributed agents.

In summary, the urgent need for early detection of suicidal ideation for suicide prevention is paramount. This project employs content analysis, feature engineering, and deep learning techniques, including deep neural networks, attentive relation networks, and federated transfer learning. The aim is to effectively detect suicidal ideation and ultimately prevent suicides, saving lives.

Publications

Suicidal Ideation Detection in Online Social Content.
Shaoxiong Ji.
Master of Philosophy, The University of Queensland. 2020.
Suicidal Ideation Detection: A Review of Machine Learning Methods and Applications.
Shaoxiong Ji, Shirui Pan, Xue Li, Erik Cambria, Guodong Long, and Zi Huang.
IEEE Transactions on Computational Social Systems, 2021.
Suicidal Ideation and Mental Disorder Detection with Attentive Relation Networks.
Shaoxiong Ji, Xue Li, Zi Huang, and Erik Cambria.
Neural Computing and Applications, 2021.
Knowledge Transferring via Model Aggregation for Online Social Care.
Shaoxiong Ji, Guodong Long, Shirui Pan, Tianqing Zhu, Jing Jiang, Sen Wang, and Xue Li.
arXiv preprint arXiv:1905.07665, 2019.
Detecting Suicidal Ideation with Data Protection in Online Communities.
Shaoxiong Ji, Guodong Long, Shirui Pan, Tianqing Zhu, Jing Jiang, and Sen Wang.
24th International Conference on Database Systems for Advanced Applications (DASFAA), 2019.
Supervised Learning for Suicidal Ideation Detection in Online User Content.
Shaoxiong Ji, Celina Ping Yu, Sai-fu Fung, Shirui Pan, and Guodong Long.
Complexity, 2018. [Code]

Feature Photo by Dan Meyers