Distant Supervision for Mental Health Management in Social Media: Suicide Risk Classification System Development Study

• DOI: 10.2196/26119
• Publication Date: 2021-08-26T14:17:16Z
• Authors: Guanghui Fu; Changwei Song; Jianqiang Li; Yue Ma; Pan Chen; Ruiqian Wang; Bing Xiang Yang; Zhisheng Huang
• Keywords: Suicide Prevention; Original Paper; Computer applications to medicine. Medical informatics; R858-859.7; Deep learning; 02 engineering and technology; Crisis prevention; Distant supervision; 01 natural sciences; 3. Good health; Mental Health; SDG 3 - Good Health and Well-being; Mental Recall; 0202 electrical engineering, electronic engineering, information engineering; Humans; Mental health; Public aspects of medicine; RA1-1270; Social Media; 0105 earth and related environmental sciences

Background Web-based social media provides common people with a platform to express their emotions conveniently and anonymously. There have been nearly 2 million messages in particular Chinese data source, several thousands more are generated each day. Therefore, it has become impossible analyze these manually. However, identified as an important source for the prevention of suicide related depression disorder. Objective We proposed this paper distant supervision approach developing system that can automatically identify textual comments indicative high risk. Methods To avoid expensive manual annotations, we used knowledge graph method produce approximate annotations supervision, which provided basis deep learning architecture was built refined by interactions psychology experts. were three annotation levels, follows: free (zero cost), easy (by students), hard experts). Results Our evaluated accordingly showed its performance at level promising. By combining our features from user blogs, obtained precision 80.75%, recall 75.41%, F1 score 77.98% hardest test data. Conclusions In paper, develop automatic classify low risk based on comments. The model therefore provide volunteers early warnings prevent users committing suicide.