Suicide Tendency Detection Method Based on Temporal-Aware Logic

Suicide is a small number of individuals who suffer from long-term and complex psychological depression and take extreme measures to end their personal lives voluntarily. The rapid development of social media platforms has prompted a large number of users to share their mental health status online. Although some progress has been made in the evaluation of suicide detection algorithms in the comments of social media users, it is necessary to combine the development history of users' mental health status to predict the changes in users' suicidal psychology accurately. Current detection algorithms mainly use attention mechanisms to evaluate the probability of word-level suicidal tendencies in user comments. However, word-level attention mechanisms may not be able to find related suicidal tendencies posts and lack the historical process of capturing psychological information transformation in user comments. Therefore, this paper introduces a novel temporal sensing module, that adaptively adjusts the temporal interval nodes of user historical comment information by analyzing the temporal logic of user information, and comprehensively detects the suicidal tendency of user comments. In implementing the algorithm, the user comment post is decomposed into multiple independent hidden states through the time series awareness module, and the most important time series scale is adaptively selected by combining the time series context information of the post. Finally, the experimental comparison in the public dataset Reddit proves that the accuracy of the proposed algorithm in detecting the psychological suicide tendency of users is better than the current baseline model.